The fix was made without incident or protest...I'm shockingly surprised. Thank you WIS.

bump - why isnt this pinned so it will stay at the top?

So you'll stick around each week, if only to bump this thread?

lol.

Hey! I can post again!

Welcome back from the time-out chair.

sample size consideration:

what if Im in a playoff race going into the last game tied for 6th in my conference and Im playing the worst team in the league who Ive beaten 6 times -all I have to do is win this last game and Im in

is the sample size too small?

Slymonium you are a PDX Bada$$. With a nasty potty mouth at times. Are you and shapandrew7 related LOL?

Quote: Originally Posted By chester_o on 10/15/2009

sample size consideration:

what if Im in a playoff race going into the last game tied for 6th in my conference and Im playing the worst team in the league who Ive beaten 6 times -all I have to do is win this last game and Im in

is the sample size too small?

chester this touches on the gist of the problem with the direction admin decided to take several years ago when matters of sample size became an on-going concern - they went the way of the simulator and statistical analysis wherein the point of the engine is to project likely outcomes using existing stat tables - the idea being they can sim thousands of match-ups thousands of times to project who will win the NCAA tournament for instance

that's all fine and well and you can draw your own conclusions about the success of that project

but the question seems to be how does the engine conceived of in this way perform for you if you are a guy who wants to run a team in a competitive league against other owners in order to see who has the best team and the answer for a lot of people including myself seems to be 'not so hot' - hence the large decline in users/teams bought and other factors relating to your other thread vis leagues not filling up

one thing admin might do to answer their own (really they say it themselves) damning criticism of their game (i.e. their acknowledgement that a single game or even 10 games is insufficient sampling) is to increase the sample size represented by each game - perhaps they could sim each game 10 times and average out the results - however I expect there might be some concerns regarding processing time and capability in this regard if they had to do this for every game in every league twice a day - maybe, maybe not - I dont know

but there you have it constructive feedback and an idea that might address the concern ( I will now prepare myself to be ignored)

My thought is instead of simulating each game 10 times ... Simulate the outcome of each play 100 times and then have to outcome be the average/most likely to occur event to happen for that play. I want to see more consistently reliable outcomes to plays and games.

There will always be crazy outliers no matter what method(s) of simulating outcomes of games or plays is/are. But the current NBA SIM engine seems to generate a much greater, than what I consider to be reasonable, amount of outliers.

When home teams shoot like 4 free throws and away teams shoot over 25 something is way out of whack.

I think the complaints about randomness distract from the real problems with the sim.

Chester,

No, beating another team six times in a row should not guarantee that you are going to beat them a seventh time. Perhaps your team should win on average 90% of the time, but that still means they should lose 10% of the time. I don't see how this particular complaint has anything to do with sample size. You are talking about one game! And who's to say that a loss in this seventh game is the aberration? Perhaps you were lucky to have won all six prior games.

I know Monkee's got an axe to grind on this particular issue (randomness), but I remain unconvinced that it is a problem with the sim. Rather, the issue is that the sim has become too predictable. There is currently only one real means of winning with the current salary structure - high efficiency, high volume three point shooting. The presence of "cookie cutter" teams is fairly convincing proof that the problem is not too much randomness.

The real problems with the sim are the salary structure that is too cheap and that overvalues certain stats relative to their impact inside the actual game, the lack of impact of defense in the game, the defensive positioning mechanism that just doesn't work, and the continuing loophole that allows players to overplay their minutes with minor if any consequences, particularly in the playoffs. Fix these, and the sim becomes playable again. "Fix" randomness, and you only exacerbate the real problems.

Quote: Originally Posted By naboimp on 10/16/2009

I think the complaints about randomness distract from the real problems with the sim.

Chester,

No, beating another team six times in a row should not guarantee that you are going to beat them a seventh time. Perhaps your team should win on average 90% of the time, but that still means they should lose 10% of the time. I don't see how this particular complaint has anything to do with sample size. You are talking about one game! And who's to say that a loss in this seventh game is the aberration? Perhaps you were lucky to have won all six prior games.

you're flirting with the gambler's fallacy here I think - I mentioned the 6 wins and overall records to illustrate that one team is presumably better than the other - but the previous 6 wins should have no real influence on the outcome of future games what should have influence is that one team is built better than the other

as for randomness you do have a point regarding cookie cutters but that doesnt change the fact that admin itself says the one game or even ten games is insufficient sample size for stat normalization - that much is a fact - can you live with it for 82 games? maybe but what about the implication for 7 game series once you get into the post season?

Quote: Originally posted by chester_o on 10/16/2009 admin itself says the one game or even ten games is insufficient sample size for stat normalization - that much is a fact - can you live with it for 82 games? maybe but what about the implication for 7 game series once you get into the post season? 

This is the primary issue with the sim to me.
If 10 games is too small a sample size, how the hell can you run a sim where championships are decided by seven-game series? The better team doesn't always win IRL, but the better team usually wins. In this sim, it's almost a crapshoot when you get to the playoffs.

I'm going to ask a really stupid question, then... How do you know which is the better team? If the team you think is better loses in a seven game series, what are the indicators of statistical outliers - as opposed to you (or any of us) just having formed the wrong assumptions about the likely winner?

The team with superior efg%, advanced stats, defense, ftas, pf, tovs and tov% without wasted minutes and possessions is generally the better team.

And I feel pretty confident that there are at least a good 20 other owners out there that can accurately judge a team vs team assessment on those categories and predict who should win a series.