GSL Site Improvements/Updates

New thread for improvements and updates to the GSL website that I'm working on.

Currently I am working on getting live scoring added to the website, which will occur on a "Game Center" page, which will display any games that are currently live.

Step one was generating box scores. Since I have been able to find NFL play by play data, instead of generating box scores by hand, I'm now matching up final scores to actual NFL games, and using the box scores of those games to match the GSL game's final score.

The box outputs on my end look like this:

Using score from game BUF@NYJ 10/26/14

Score Summary (BAL @ LV):

Quarter 1
[0-7] 09:42 - LV touchdown
[0-14] 04:40 - LV touchdown

Quarter 2
[7-14] 14:25 - BAL touchdown
[7-21] 11:56 - LV touchdown
[7-24] 06:21 - LV field goal
[14-24] 01:55 - BAL touchdown
[17-24] 00:04 - BAL field goal

Quarter 3
[17-27] 08:34 - LV field goal
[17-30] 01:18 - LV field goal

Quarter 4
[17-37] 14:16 - LV touchdown
[17-43] 09:56 - LV touchdown (extra point attempt failed)
[23-43] 04:11 - BAL touchdown (two point conversion attempt failed)

FINAL
BAL - 23
LV - 43


Since I have the full play by play for the real NFL game (BUF@NYJ 10/26/14), I should be able to just walk through each play in the real game, and show on the website the quarter, time remaining, team with possession, where the ball is, down and distance.

That'll be some work to put together, but once I think I'm in a good place, I'm going to schedule an exhibition game, with a GSL 2022 expansion team, the Utah Trappers, owned by Andrew Seagraves, being the home team.

I'll continue scheduling the Trappers in exhibitions games each week until I have live game data figured out, then I'll make all GSL games show when live in the game center. Hopefully by mid-season we will have live play by play for the entire GSL!

Eventually I can expand on the live play by play, by producing game stats and play descriptions. This is likely something for the 2022 season.

This is a fantastic idea. Let's hope the dice don't land on any scoragami

rbronish wrote:This is a fantastic idea. Let's hope the dice don't land on any scoragami

Well, the way around that is, there is almost always a score very close or similar to a scoragami. I can use a score that's close, and modify it, often by just one play, to get the score I want.

For arguments sake, lets say a simulated score ends up 43-12, and it's a scorigami. There's no NFL game that ended 43-12, but, there might be games that ended 43-13, or 42-12, where the difference is an extra point. I can easily adapt the real life game by making a team miss an extra point, or make an extra point that in real life was missed. Same goes for field goals. I can always take away a field goal from a team, or add one in a place where a team missed. Stuff like that lets me get any score I need from an actual NFL game.

It seems like I've got a small bug in the standings code.

Head to head victories don't seem to be counting towards the standings as they are supposed to. For instance, In the Central division, the Diesel just beat the Comets, and with both teams in a two team tie, the tie breaker goes to the Diesel, but the Comets are still showing ahead of the Diesel. I will need to fix this.

I should note that the conference standings are correct, with the Comets ahead of the Diesel in the conference, because of the three way tie with Seattle.

Standings code has been updated. The issue was based on me looking for two teams tied for the entire league, rather than on an individual conference or division.

The correct league standings are now showing.

If at any point during the season, something in the standings does not look correct, please let me know. At the end of the season I will audit the standings and calculate them by hand if need be, to make sure the correct teams are in the playoffs.

I believe we should be a spot above Miami, since we have the same record + we won the H2H matchup in week 1.

NeauXone wrote:I believe we should be a spot above Miami, since we have the same record + we won the H2H matchup in week 1.

The head-to-head only applies in 2-way ties; since there's a 5-way tie in the conference, it instead goes to conference record, in which Miami is .500 and Louisville is .333.

Untitled0717 wrote:

NeauXone wrote:I believe we should be a spot above Miami, since we have the same record + we won the H2H matchup in week 1.

The head-to-head only applies in 2-way ties; since there's a 5-way tie in the conference, it instead goes to conference record, in which Miami is .500 and Louisville is .333.

Okay, that makes sense. Didn't realize that.

This may not apply here, but I know with the NFL if you have multiple teams in multiple divisions, you apply in the in-division tiebreakers first and then apply appropriate conference tiebreakers with only the top team in each division..     so after you sort the Jacksonville-Louisville tiebreak and the Miami-Baltimore-Norfolk tiebreak, then you would compare Wisconsin-Louisville-Baltimore for first Wild Card, and so...

At least that's how the NFL did it at one point, but I'm not sure if it still is.

Rob Holecko wrote:This may not apply here, but I know with the NFL if you have multiple teams in multiple divisions, you apply in the in-division tiebreakers first and then apply appropriate conference tiebreakers with only the top team in each division..     so after you sort the Jacksonville-Louisville tiebreak and the Miami-Baltimore-Norfolk tiebreak, then you would compare Wisconsin-Louisville-Baltimore for first Wild Card, and so...

At least that's how the NFL did it at one point, but I'm not sure if it still is.

Yes, that's how the NFL still does it. I wasn't sure how to code that at the beginning of the year, so because that wasn't in the official rules for this season, I don't know that I want to change that.

Maybe we can hold a league vote on adding that into the standings calculations. That's something that I can add in for next year for sure though.

A simple improvement I'd suggest is an expanded standings page with things like Strength of Victory, Home/Road wins, etc.

92Mafia wrote:A simple improvement I'd suggest is an expanded standings page with things like Strength of Victory, Home/Road wins, etc.

That'll be something to add for the 2022 season. I don't factor in strength of victory or schedule into standings this year, but would like to next year. Thing is, with this season, a lot of teams play the same teams. For example, each conference plays every team in their conference. Next year, with more teams, that won't be the case.

Only other thing I'll have to think about with adding those things, is, there's no room on the page to fit the actual standings data in if I add more data. So, I'll probably just add strength of victory, and maybe strength of schedule.

With 32 teams, I can make the tie breaking procedures the same as the NFL's.

In order for me to get live games to work, I need to update the format of the game timestamps in the database.

This could possibly break the code that checks to see when a game should be simulated and have it's score posted on the website.

So, today, game scores may show up just after 3 hours after the games complete, since I will be kicking off the simulations manually through a little script I have. Just be aware. If all goes well, the games will still simulate automatically.

Made a good breakthrough today with the live updates. I just need to tie my test randomized scores to the actual simulation.

This is my current output, with some test data and randomized scores (the SA-CIN game data here has nothing to do with todays game and final score, it's al part of testing)


Live Game: 2021_SA-CIN^5

| *SA - 0 | CIN - 0 | 1st 2:58 | 1st & 10 | Ball On SA 34 |

Last play: Punt


Live Game: 2021_GSL-SL^7

| *GSL - 0 | SL - 7 | 1st 9:32 | 3rd & 10 | Ball On GSL 27 |

Last play: Pass play




I'll also update the formatting on how the data comes out, so it looks nice. The Score summary and box score is also something I store, as well as all the game play by play. Eventually that will be something that could be shown as well.

I have a tentative Exhibition game now scheduled for Tuesday evening, to do a first run of a live game.
The play details won't be very informational, but there will be live down and distance data, as well as scoring of course.

I can hopefully fix any bugs I run into before the next exhibition game on Saturday March 27th.
If all goes well with that game, the week 6 Monday night game between Los Angeles and St. Louis will have live updates.
And, finally, if that game goes well, then all of week 7 onward could have live game updates.

That's my hopeful schedule, but don't be surprised if I get delayed. I'd hope for sure by week 10 we can have all the games live.

I see Tuesday's exhibition was moved to right now, is there any way we can see the live updates?

NeauXone wrote:I see Tuesday's exhibition was moved to right now, is there any way we can see the live updates?

Sorry, I was testing live updates (which look pretty good):

Pay no attention to that score, this game will be simulated tomorrow.

I moved this game back to its originally scheduled time.
I need to work on adding a link to go to live games on the normal weekly page. Tomorrow the game will be live for real, just for now I was running tests. Other than that, the only thing that hasn't been working correctly is the length of half time. Should be 12 minutes.

*EDIT
Looks like I also forgot a way to show possession.

It looks like in the Western conference standings, that the Comets are somehow ahead fo the Diesel again, which is incorrect.
I'm going to have to figure out why that is. Those two teams are the only ones in an exact tie with each other in the conference in their positioning.
I think at this point, I will have to default back to division standing preceding conference standings.

This will effect 3+ team tie breaks. It will ensure that a team in a higher division standing will have a higher conference standing. I probably won't get to fixing this until later in the week.


For now, my big goal today is making sure live updates work. The game time for tonights game might jump around throughout the day, but the official time for tonight's game is 6:30PM EDT.

So it seems like the problem with the scores is related to the score not updating in the database at the end of the game.

I think I have a way to solve that though, and I'll keep testing today.

*** EDIT ***

Okay, I think this issue is finally solved, but I guess we will see Thursday night.
I've also gotten the game times cut down to a little over 3 hours by shortening time between plays where there's stoppages.

Adapting the Gridiron Fields Database to a GSL Fields Database turned out to be easier than I thought:
http://gridiron-uniforms.com/GSL/controller/controller.php?action=fields-home

Austin Snelick wrote:Adapting the Gridiron Fields Database to a GSL Fields Database turned out to be easier than I thought:
http://gridiron-uniforms.com/GSL/controller/controller.php?action=fields-home

Thanks, Austin. I always love additions to the field databases lol.

As of today, gridiron-uniforms.com has an SSL certificate, therefor all http requests redirect to https. This should help with browser support and security going into the future.

Currently, the standings have Las Vegas ahead of St. Louis, despite St. Louis having the better conference record. Also, the Express PF and Brewers PA are off by 1; I'm assuming the standings didn't update with the LA-STL score was fixed.

Simulations are running smoothly today, also quick question for Austin: Have you considered having a Ticker running across the top of the site that shows scores?

Untitled0717 wrote:Currently, the standings have Las Vegas ahead of St. Louis, despite St. Louis having the better conference record. Also, the Express PF and Brewers PA are off by 1; I'm assuming the standings didn't update with the LA-STL score was fixed.

Yes, I see that too, along with a couple other problems in the standings. I'm going to have to rewrite most of the standings code to get it to work right, but maybe for now, I can post on Tuesday's the official standings until I have time to get it fixed.

NeauXone wrote:Simulations are running smoothly today, also quick question for Austin: Have you considered having a Ticker running across the top of the site that shows scores?

I don't think I'll add a ticker, just because the scores are pretty accessible from the main page of the site, and I should be focusing my work on other things at the moment. Maybe that's a down the road thing.

Took me a while to get around to adding this, but I've added play by play data to the single matchups page.
I've also added a clear link to the single matchups page, labeled as "Box Score", to make it more clear for the games that do have box score data.

So I've been thinking recently about the future of the GSL, and how things would progress from season to season, and it kept coming up in my head... how could we possibly show natural growth or decline of teams even though the ratings are completely random each season? And over the past few weeks I've been thinking of ways this could be achieved, showing a level of consistency between seasons while also giving every team a chance.

One idea I came up with was developing some sort of equation similar to the betting lines that could be used to push out new ratings for each team, that could be somewhere in the ballpark of said team's previous rating from the season before. This way, teams don't go from 0-18 to 18-0 caliber, but still display that said team isn't the same squad they were the season before. I'm not sure what the exact numbers for this would be but I think this method would work.
I think that this would work especially well because it would make teams better/worse based on their primary strength. For example, Miami or Jacksonville next season could potentially build off of their already great defenses, or add offensive pieces to complement their defense and become good teams. Pennsylvania could lose offensive firepower and still be a great team, but not a terror squad like they are this season.

NeauXone wrote:So I've been thinking recently about the future of the GSL, and how things would progress from season to season, and it kept coming up in my head... how could we possibly show natural growth or decline of teams even though the ratings are completely random each season? And over the past few weeks I've been thinking of ways this could be achieved, showing a level of consistency between seasons while also giving every team a chance.

One idea I came up with was developing some sort of equation similar to the betting lines that could be used to push out new ratings for each team, that could be somewhere in the ballpark of said team's previous rating from the season before. This way, teams don't go from 0-18 to 18-0 caliber, but still display that said team isn't the same squad they were the season before. I'm not sure what the exact numbers for this would be but I think this method would work.
I think that this would work especially well because it would make teams better/worse based on their primary strength. For example, Miami or Jacksonville next season could potentially build off of their already great defenses, or add offensive pieces to complement their defense and become good teams. Pennsylvania could lose offensive firepower and still be a great team, but not a terror squad like they are this season.

Well, I do have a method, that actually works pretty well with the way I'm currently simulating everything, which again can be based on a simple 6 sided die. I would roll for each team for both offense and defense, where rolling the following would effect the rankings accordingly:
1: -2
2: -1
3: 0
4: 0
5: +1
6: +2

So, for example, lets take a team that has offensive and defensive rankings of 3 and 4 respectively, and then I roll a 6 and a 2. That team would now have rankings of 5 and 3 for offense and defense. And, of course, 1 and 6 are limits, so you can't go below 1 or above 6, regardless of the rolls.

I think making rankings of the following year dependent on the previous year makes the league more realistic, but, given the nature of this league as a simulation league, as team owners, I would think it would be nice to know that your team has a chance every year, despite what happened the previous season. Take a team, like Atlanta for example. It can't be fun for Rob to watch his team get obliterated every week, and to know that next year, even if his team gets lucky with rankings changes, the Flying Fleet still won't be good enough to compete for title.

Now, rankings changes in the off season could vary based on where a team finished the previous year. For instance, a team at the bottom in the offseason gets a better draft pick, therefore they have a better chance of improving. Something like:
1: -1
2: 0
3: 0
4: +1
5: +2
6: +3
Where the team has a better chance of drafting a generational talent that drastically improves the team.

Like wise, the best teams may have different odds at improving. Something like:
1: -2
2: -1
3: -1
4: 0
5: +1
6: +1
Or some other combination.

At least for 2022 I'd like to reset everything, as the league expands to 32 teams. Think of it as, all players go into a draft pool to supplement a 32 team draft to support the new expansion teams.

It's possible for 2023 I could change things up and base that season on 2022 season team rankings.

I like both ideas. But with a 32 team league next year. An expansion draft would occur. As well as some moving of talent like Free Agency and trades. And in the "real world" of pro sports. Nothing is always the same anyway from year to year. Everybody looks to get better. Even a league Championship team. :-)

John

Austin Snelick wrote:

NeauXone wrote:So I've been thinking recently about the future of the GSL, and how things would progress from season to season, and it kept coming up in my head... how could we possibly show natural growth or decline of teams even though the ratings are completely random each season? And over the past few weeks I've been thinking of ways this could be achieved, showing a level of consistency between seasons while also giving every team a chance.

One idea I came up with was developing some sort of equation similar to the betting lines that could be used to push out new ratings for each team, that could be somewhere in the ballpark of said team's previous rating from the season before. This way, teams don't go from 0-18 to 18-0 caliber, but still display that said team isn't the same squad they were the season before. I'm not sure what the exact numbers for this would be but I think this method would work.
I think that this would work especially well because it would make teams better/worse based on their primary strength. For example, Miami or Jacksonville next season could potentially build off of their already great defenses, or add offensive pieces to complement their defense and become good teams. Pennsylvania could lose offensive firepower and still be a great team, but not a terror squad like they are this season.

Well, I do have a method, that actually works pretty well with the way I'm currently simulating everything, which again can be based on a simple 6 sided die. I would roll for each team for both offense and defense, where rolling the following would effect the rankings accordingly:
1: -2
2: -1
3: 0
4: 0
5: +1
6: +2

So, for example, lets take a team that has offensive and defensive rankings of 3 and 4 respectively, and then I roll a 6 and a 2. That team would now have rankings of 5 and 3 for offense and defense. And, of course, 1 and 6 are limits, so you can't go below 1 or above 6, regardless of the rolls.

I think making rankings of the following year dependent on the previous year makes the league more realistic, but, given the nature of this league as a simulation league, as team owners, I would think it would be nice to know that your team has a chance every year, despite what happened the previous season. Take a team, like Atlanta for example. It can't be fun for Rob to watch his team get obliterated every week, and to know that next year, even if his team gets lucky with rankings changes, the Flying Fleet still won't be good enough to compete for title.

Now, rankings changes in the off season could vary based on where a team finished the previous year. For instance, a team at the bottom in the offseason gets a better draft pick, therefore they have a better chance of improving. Something like:
1: -1
2: 0
3: 0
4: +1
5: +2
6: +3
Where the team has a better chance of drafting a generational talent that drastically improves the team.

Like wise, the best teams may have different odds at improving. Something like:
1: -2
2: -1
3: -1
4: 0
5: +1
6: +1
Or some other combination.

At least for 2022 I'd like to reset everything, as the league expands to 32 teams. Think of it as, all players go into a draft pool to supplement a 32 team draft to support the new expansion teams.

It's possible for 2023 I could change things up and base that season on 2022 season team rankings.

This is actually along the lines of what I was thinking! I think the method you listed there at the end works really well, it gives the teams at the bottom a chance to become contenders the next season, with room to improve via the weekly ranking changes as well.

So this is similar to what I'm doing with the UWFFL (which will be explained at great depth if I ever get caught up enough and ever put out the video explaining the UWFFL simulation process)  ... but in regards to year-to-year off season adjustment what I did was take NFL data and look at how much teams typically improved or regressed from year to year... and took a range of the middle 80% ... i.e. the middle 80% of teams had a range of from what to what (+x amount of winning percentage to -y amount of winning percentage)  and then randomized a number for each team and then would adjust their skill rating a corresponding amount, indexed to that range.  After testing it I figured 80% was about the right percentage to exclude the outliers at the high end and low end.  

Now we also have in-season migration factor based on performance that I hope would also simulate a team "getting hot" (or cold)...  Basically for this I determined what is the maximum range that a team's skill rating would "migrate" or go up or down during course of a season.... and so for instance let's say we went with 20 points then if it is a 14 game regular season then every time a team wins they increase their skill rating by 20/14th of a percentage point.   Because whenever a team wins they have outperformed their expected win % and whenever a team loses they underperform it ....    (even if the #1 team faces the worst team and the expected winning percentage might be .98 for them and .02 for the other team... a win is a 1.00 and a loss is a .00)  so basically whenever a team wins it goes up a little (20/14) and loses it goes down a little.

Under that scenario if a team goes 14-0 they will gain +20 points in migration during the course of a year and likewise a team that goes 0-14 will lose 20 points.

I hope to let you all critique the full UWFFL simulation process soon.

You may have noticed a change to the "GSL Teams" section in the header, with slots added for new teams for 2022.
With that change, I have also changed the link that each team page goes to, now going to a "Team History" page. This will track the logos, wordmarks, uniforms, and fields for each team across the seasons. (although, if a team relocates, currently, the old team/location won't show up on the history page, just the current team city/name combo).

Most teams are in their first season, but for the original 6 teams, you will see data from the 2020 season:
https://www.gridiron-uniforms.com/GSL/controller/controller.php?action=team-history&team_id=LV

This will be how I will preserve team history on the site in one location, as I'm sure you all will change some things about your teams throughout the years.
Ownership will also be shown by year, so if you decide to leave the GSL as an owner, but want someone else to take over your team, you will still be shown as the original team founder.

Austin Snelick wrote:You may have noticed a change to the "GSL Teams" section in the header, with slots added for new teams for 2022.
With that change, I have also changed the link that each team page goes to, now going to a "Team History" page. This will track the logos, wordmarks, uniforms, and fields for each team across the seasons. (although, if a team relocates, currently, the old team/location won't show up on the history page, just the current team city/name combo).

Most teams are in their first season, but for the original 6 teams, you will see data from the 2020 season:
https://www.gridiron-uniforms.com/GSL/controller/controller.php?action=team-history&team_id=LV

This will be how I will preserve team history on the site in one location, as I'm sure you all will change some things about your teams throughout the years.
Ownership will also be shown by year, so if you decide to leave the GSL as an owner, but want someone else to take over your team, you will still be shown as the original team founder.

Thanks Austin. Looks great. Much like the entries in the NFL section. And I do not plan to go any where. I love this way to much. :-)

sfcom1 wrote:

Austin Snelick wrote:You may have noticed a change to the "GSL Teams" section in the header, with slots added for new teams for 2022.
With that change, I have also changed the link that each team page goes to, now going to a "Team History" page. This will track the logos, wordmarks, uniforms, and fields for each team across the seasons. (although, if a team relocates, currently, the old team/location won't show up on the history page, just the current team city/name combo).

Most teams are in their first season, but for the original 6 teams, you will see data from the 2020 season:
https://www.gridiron-uniforms.com/GSL/controller/controller.php?action=team-history&team_id=LV

This will be how I will preserve team history on the site in one location, as I'm sure you all will change some things about your teams throughout the years.
Ownership will also be shown by year, so if you decide to leave the GSL as an owner, but want someone else to take over your team, you will still be shown as the original team founder.

Thanks Austin. Looks great. Much like the entries in the NFL section. And I do not plan to go any where. I love this way to much. :-)

Agreed, this experience has been incredile so far.
Do you think eventually you'll add the special tabs from the GUD, like the color v. color and white at home pages?

NeauXone wrote:

sfcom1 wrote:

Austin Snelick wrote:You may have noticed a change to the "GSL Teams" section in the header, with slots added for new teams for 2022.
With that change, I have also changed the link that each team page goes to, now going to a "Team History" page. This will track the logos, wordmarks, uniforms, and fields for each team across the seasons. (although, if a team relocates, currently, the old team/location won't show up on the history page, just the current team city/name combo).

Most teams are in their first season, but for the original 6 teams, you will see data from the 2020 season:
https://www.gridiron-uniforms.com/GSL/controller/controller.php?action=team-history&team_id=LV

This will be how I will preserve team history on the site in one location, as I'm sure you all will change some things about your teams throughout the years.
Ownership will also be shown by year, so if you decide to leave the GSL as an owner, but want someone else to take over your team, you will still be shown as the original team founder.

Thanks Austin. Looks great. Much like the entries in the NFL section. And I do not plan to go any where. I love this way to much. :-)

Agreed, this experience has been incredile so far.
Do you think eventually you'll add the special tabs from the GUD, like the color v. color and white at home pages?

Yeah, that was something I was planning to get around to eventually. I'm hoping it's mostly just a copy and paste kind of thing, where I change some paths and variable names to fit the GSL. The database queries wouldn't have to change I don't think. The biggest task there is setting up the little script on my side where I can select all the color uniforms and have them be marked as color in the database.

I still haven't gotten around to fixing the standings code. It's much more work that I anticipated (the calculation is pretty complex, so I'd basically need to rewrite how I do standings entirely to do it right).

For sanities sake, the Toronto Terror are the 4th seed in the Eastern Conference, ahead of the Cincinnati Knights, who would then be the 5th seed. This is based on division standings, where the Terror have a better division record, with both teams tied in head to head matchups.

I'm currently working on fixing the conference standings sections. So you may see some errors or weird things in the standings section while I'm working on that.

Standings are now (finally) correct.

Toronto is now showing as 4th in the Eastern Conference, ahead of Cincinnati, as Toronto has the division tie breaker.

Other sorting in the Eastern conference.
Washington must be ahead of Baltimore, based on division tie breaker.
Wisconsin won the head to head matchup against Washington, therefore Wisconsin is ahead of Washington.
This creates the order of Wisconsin, Washington, Baltimore in the Eastern conference standings.

Lets still keep an eye on the standings over the next few weeks, to make sure they are correct.

Now that we have settled on "Redemption Game" as the conference runner up game name before the Gridiron Bowl, I have added that game to the schedule on the website. Although this years game is called the "2021 Playoff Bowl", next year the game will be called the "2022 Redemption Game". I've coded the game in the database as Redemption Game, including this year for simplicity.

Royals/Caps are breaking the 4th wall. Last plays are Rams/Seahawks

rbronish wrote:Royals/Caps are breaking the 4th wall. Last plays are Rams/Seahawks

It seems the Royals-Captains game is using the same NFL game as the Brewers-Sockeyes game, and is still using STL and SEA in the game center. This has happened a few times in the past with other games.

Untitled0717 wrote:

rbronish wrote:Royals/Caps are breaking the 4th wall. Last plays are Rams/Seahawks

It seems the Royals-Captains game is using the same NFL game as the Brewers-Sockeyes game, and is still using STL and SEA in the game center. This has happened a few times in the past with other games.

Yeah, I've noticed this for a couple games. I'm not sure what the cause is, but probably bad NFL data that didn't transfer properly. Also, we were bound to repeat games, since I only have so many NFL games to choose from, sometimes with only one instance of a game score, so if we hit that score multiple times, it would pull from the same game multiple times.

I'm really hoping that I have time during the offseason to revamp the simulation a little bit. It'll be hard to account for team rankings and do a play by play simulation, so my thoughts were, I run a bunch of game simulations, and back log a ton of games with play by play data. Since, what I do now is simulate a game, then pick the NFL game with the same final score to get the play by play, I could do a similar thing, just with games I've simulated, and pull play by play form those games. The hard part is writing the simulation though, because it's hard to make coaching decisions. I'm thinking, the AI coaching decisions will just be based on actual NFL game data, and what a coach did in the same down/distance/time remaining. It certainly won't be perfect the first go around, but the key is that the simulation is done before the play by play, and the play by play, no matter what happens (errors or whatever) don't impact the outcomes of games, because that is predetermined.

Austin Snelick wrote:I could do a similar thing, just with games I've simulated, and pull play by play form those games. The hard part is writing the simulation though, because it's hard to make coaching decisions. I'm thinking, the AI coaching decisions will just be based on actual NFL game data, and what a coach did in the same down/distance/time remaining. It certainly won't be perfect the first go around, but the key is that the simulation is done before the play by play, and the play by play, no matter what happens (errors or whatever) don't impact the outcomes of games, because that is predetermined.

Just as long as you do not pull from the Detroit Lions for the past say 30 years. As the coaches have always been bad. lol

LA-ATL: a tank bowl so strong it bends the fabric of the game itself.

That's certainly an anomaly.

The simulated final score is in the database. I want to wait for other games to finish before making score adjustments (if needed) so I don't disturb other game simulations.
Hopefully this is not an issue next year. Maybe for the playoffs, I'll pre-simulate games, then manually choose an NFL game to match play by play, so we don't have these kinds of things happen then.

I can confirm that the final score of ATL 21 - LA 20 is the correct, and simulated final score.

Austin Snelick wrote:I can confirm that the final score of ATL 21 - LA 20 is the correct, and simulated final score.

Wait, I'm confused.... we won? Are you sure? It doesn't seem like us.

Rob Holecko wrote:

Austin Snelick wrote:I can confirm that the final score of ATL 21 - LA 20 is the correct, and simulated final score.

Wait,  I'm confused....  we won?    Are you sure?   It doesn't seem like us.

Ha, yeah, the seems unreal that Atlanta would win, but it did happen, even if one of the score boards said it didn't happen.

But anyways, what had happened was, the NFL data I had for that game marked the extra point (in this case, Los Angeles' game tying score) as in the 5th quarter, or not in the fourth. Don't know why, but that data isn't very good. So, the score at the "end of the game" picked up by my code, wasn't the actual end of the game score, because of the data.

What I'm going to have to do for next year, as step one into building a play by play simulation, is use the NFL data I have now, to simulate some full games to get a pool of non NFL games, that I simulated with a bunch of scores that can be used to choose from for games next year. I'll have to break dow a whole mess of game situations though. What I really need to do is get another python programmer on board to help me build the simulation.

But how did it reverse the teams ... according to the summary it looked like Atlanta was ahead 20-13 and LA scored TD + 2 pt conversion? Just curious, we'll definitely take the win!

Rob Holecko wrote:But how did it reverse the teams ... according to the summary it looked like Atlanta was ahead 20-13 and LA scored TD + 2 pt conversion?   Just curious,  we'll definitely take the win!

It might take some more digging into what actually happened, but all I know is, I have a database table that tells me what the final simulated score of the game is, and that's inserted into the database before the code even makes the selection of the NFL play by play data. So that database table with the final simulated score is the official game score.

But thank goodness this happened on a game that didn't mean anything, rather than a big game like the Sunday night game. I should start throwing out the NFL games with bad play by play data, so I minimize problems moving forward.