2010 RULES: Playoff Updates

[Goodell]

We'll start the off-season discussions off going backwards with something fresh in my mind just coming off the playoffs. This gets a little technical so if you aren't sure don't hesitate in asking.

We changed how we did playoff updates this year. Unfortunately the planned change looked good on paper but was a little too overwhelming with an extra league also. It got to the point where each team would take around 45-60 minutes to update (12 total teams championship round, 6 total teams last round = nearly 18 hours of stats crunching the last two rounds). The league needs to be somewhat fun for me and wasn't as the playoffs went on and more work piled up to get the next batch of recalculated stats ready.

I actually kind of liked the updates being a little different each game because that's reality. You don't get identical performances over and over from round to round. QB gets 2 TDs and 0 INT one round, then 1 TD and 1 INT the next, etc. And all more fair than before where a player could either really benefit or get screwed on rounding issues where a .4 TD per game would never have had a TD in his update in previous years (.5 rounds up to 1) but this season would in some of the games with a new way of figuring the average over multiple games.

Here's how it was done this year (discussed last off-season).

[*] Player X stats calculated on an individual per game basis.
[*] Multiply those averages by individual health status (either .99, .85 or .75 depending) if not healthy.
[*] Multiply those by the number of playoff games he'd have played this round (for all super bowl teams I believe it would be the 3rd playoff game for each of their players).
[*] Go back and total up stats from previous averaged updates from previous playoff rounds.
[*] Give Player X the difference remaining for him to reach his projected stats for X number of games.

We ran into some trouble when it looked like teams who played the first round and now going against a team off a bye might get statistical advantages and perhaps have more upsets in the second round in some cases mostly due to the lower seeded teams having played more games and their averages calculated differently. So I put that off a round until all teams had played multiple games and had their averages calculated the same way for both. For example, with my team my QB would have got an extra TD in the second round and maybe a better chance of winning against a higher ranked team having to use a different average method for their players since it was their first playoff game and not over multiple games as mine.

But it took more manual labor going back and looking at each individual player and figuring their multi-game average than we can continue going forward. I eventually got a spreadsheet where it did more of that work for me, but still was taking too long and if adding more leagues in the future won't be managable.

But it's possible that we could keep something similar to this multi-game whole number average that changes slightly from round to round (2 TDs one round, 1 TD the next, etc.) but have me automate that much more to reduce the labor, or perhaps move to something entirely different in the future like using average fractions for each and every round (1.4 TD / .8 INT) and adjusting the sim to not have to use whole numbers as you get with real games but accept fractions.

If moving to fractions, it would be much easier to upload one update for each player that doesn't change, but the sim currently doesn't deal well with fractions and much of the programming is based upon if there is a TD or fumble or INT, etc. left in the update. It would most likely be efficiently implemented by the system tossing a dice at the start to see if that 1.4 TD was going to be 1 or 2 TD in the update (like 40% chance it would give a 2 TD update and 60% chance for 1 TD update). But would leave much more up to randomness than how we do now. But it could possibly also look at the last game update saved and see if that extra TD in last game to not have it this game or something. It could really adjust the script a lot and stick with fractions throughout, but might be a lot of work that isn't relevant at all except for playoffs and byes as you'd get whole numbers in the vast majority of real game updates. Also if you keep throwing a 40% chance of something over and over and over in play after play after play, it may end up always happening eventually in a game with so many chances instead of just being a real 40% chance.

DEFAULT: After the playoffs this year, I don't really want to do it the same way again and the league probably doesn't want more chances for less activity as experienced then as I fought doing it each day. I'd personally lean now toward having unchanging fractions as part of the update and trying to reduce the randomness that comes with that by looking back at previous game updates in the playoffs.

We can get into more details of next year's playoff update process when a general direction seems to be emerging through the polling and discussions.

[Joe]

Gotta go with the fractions, IMO. Its easier and the randomness would make things a little more interesting!

[Goodell]

I also wanted to mention a spreadsheet that John W sent to me recently that breaks a seasonal average into different whole numbers per round. The vote isn't going that way currently, but if we stayed the way it was this season might be more tools like this to make easier.

http://www.fangm.com/football/Playoff%2 ... ulator.xls

[Royce R]

Is it possible instead of a 40% chance of that 2nd interception to do it all from pass attempts.

If a qb throws 50 passes and has 1 int on the game then it would be a 2% chance of an INT on each pass attempt.

If a qb throws 25 passes and has 1 int then it would be a 4% chance of an INT on each attempt.

I would even like that in the regular season.



If a qb throws 10 passes in RL and has 1 pick. We can throw 50 passes and have 1 pick in the sim. And backwards of that too.

I would really like to see it where the QB has a 10% chance of a pick in the above situation. Now that is on each throw. So we may see an inflated number of picks.