General Forum

some responses

Posted By: Math Boy
Date: 25 Sep 01, 3:59 pm

In Response To: bad beats versus getting beaten badly (Math Boy)

Just replying to people, and giving counterpoints...

Don said, "Guy goes 5-1. Wins all five games by one point and the one he loses is a blowout by 50! You decide he sucks! :-) "

James said, "Hmmm.... maybe the reverse procedure would be better-- assume that any huge wins/losses are outliers and measure performance just on the close games? :) Yes, I know you went 10-4 this week, sir, but since 7 of your wins were blowout games, I've adjusted your results to 3-4 and decided you are a losing handicapper..."

We never conclude someone sucks or is great on the basis of 6 games anyway. Otherwise goat would be the man. :) But if a guy does win all 5 games by one point, you know much luck is involved. Small margin games could easily go the other way on any of a number of plays just by having different referees. I would think someone is a better handicapper if they go 5-1 and win their bets by an average of 14 points rather than just scraping by on every win.

It may make sense to have some function whereby you censor the distributions based on the number of games. So if you are only looking at 5 games, you would call anything lower than a 15% result a 15% result. If you are looking at 500 games, you call anything lower than a 1% result a 1% result.

James said, "If a successful handicapper is aiming for a success rate of roughly 55%, those 1-2 point wins are EXACTLY the edge that he is looking for. That is, a successful handicapper is primarily looking for small mistakes in the line, especially around key numbers. Find a +3.5 line that you think should be +2.5, and that's a huge edge. Find a -8.5 line that you think should be -30? Shouldn't happen too often."

The difference between +2.5 and +3.5 may be only one point, but that is why I said to use the probability distribution instead of the points. That one point difference would be a difference of about 15% on the probability distribution.

Those one two point wins may be exactly the edge the handicapper is looking for in the long run. But do you really believe that someone is going to win all their games by 2 points and lose all their games by 20 points? You are saying that you may rarely find a -8 that should be -30. But on the converse side, if you think a -30 should be -8, and you are wrong, there is something seriously wrong with your handicapping model.

StevieY said, "When you take a game that is only played once, many things can happen. Using the OSU/UCLA game as an example, both teams had plenty of chances to score. There were mistakes left and right, and a few missed kicks. It was a very sloppy game. Play that same game another 99 times, I doubt you would get as badly played game again. Yeah it wound up 13-6, and I will take my win, but it could have easily been a higher scoring game with what happened. That's the flaw, especially in college, things can suddenly snowball(sometimes for no reason) and make it "abnormal".

If I lose a game, I'd rather lose it by 50 than by 1, no agonizing.:)"

Don also said, "So long as wagers are decided by winning or losing to the spread -- and not by action points -- I don't think most of us care if we win a game by one or 41. And we shouldn't care about the loss, either."

James also said, "When UCLA/OSU score only 20 points when lined at 50, I don't think many of us assume that in a hypothetical ensemble of "identical" games, the mean total would be 20. If the "true" mean total of this ensemble should be 45 or 42 instead of 50, the 20 point total is still an outlier. Yes, the total of 20 is more likely to come from a distribution centered at 42 than 50, but it's still a low probability event."

Each game is only played once, but we are adding up many games for the overall distribution. We are relying on regression to the mean for our estimate. But if we did lose a game by 50 points, we have three choices for what happened. Either the opposing team got lucky, our model was flawed in some respect, or both! For example, Denver was underated by anyone who bet on them in both of their first two games. That was a flaw in the model that has to be adjusted. But Denver would not beat Arizona 21 points if they had to play that game 10 times. We look at several picks to remove the "if they had played 100 times" effect.

Math Boy

Messages In This Thread

bad beats versus getting beaten badly -- Math Boy -- 25 Sep 01, 11:23 am

Obvious downfall -- Don Schlesinger -- 25 Sep 01, 1:44 pm

not so obvious -- Dunbar -- 26 Sep 01, 6:28 am

To me, a waste of time -- Don Schlesinger -- 26 Sep 01, 1:23 pm

an observation -- Dunbar -- 1 Oct 01, 11:55 am

My great claim to fame :-) -- Don Schlesinger -- 1 Oct 01, 5:26 pm

I'm definitely impressed! -- Dunbar -- 2 Oct 01, 7:18 am

Jr. Mets, I believe -- Don Schlesinger -- 2 Oct 01, 11:51 am

who was the handicapper nt (nt) -- steve -- 25 Sep 01, 2:14 pm

Contrariwise... -- James -- 25 Sep 01, 3:03 pm

One problem -- StevieY -- 25 Sep 01, 3:16 pm

selection bias? -- Lefty -- 25 Sep 01, 3:46 pm

some responses -- Math Boy -- 25 Sep 01, 3:59 pm

close games -- Vetsen -- 25 Sep 01, 4:39 pm

How to be sure?? -- Don Schlesinger -- 25 Sep 01, 5:04 pm

I miswrote -- Math Boy -- 25 Sep 01, 6:58 pm

Math Boy, I AM the man! -- Goats -- 26 Sep 01, 7:29 am

I somewhat agree (and my results)... -- MDMAniac -- 25 Sep 01, 6:11 pm

I strongly agree with your post. -- fezzik -- 25 Sep 01, 9:05 pm

More results (and commentary) -- Colin Caster -- 25 Sep 01, 10:55 pm

IMO, your post was the best in a great thread -- fezzik -- 26 Sep 01, 7:29 pm

I have a different take on this thought provoking question: -- samo666 -- 26 Sep 01, 6:11 am

I like your example -- Bad Cutter -- 26 Sep 01, 8:13 am

To me, . . . -- Don Schlesinger -- 26 Sep 01, 1:29 pm

Excellent post. You may be correct. -- samo666 -- 26 Sep 01, 2:20 pm

The reason I think it's important... -- MDMAniac -- 26 Sep 01, 2:58 pm

Being better than the line -- StevieY -- 26 Sep 01, 4:23 pm

I'd like to meet you one day, Stevie -- Don Schlesinger -- 26 Sep 01, 5:34 pm

It would be my honor (nt) -- StevieY -- 27 Sep 01, 7:50 pm

Typo -- Don Schlesinger -- 27 Sep 01, 8:37 pm

You gotta learn to blame that on the keyboard (nt) -- Razor -- 28 Sep 01, 5:46 am

Right! :-) -- Don Schlesinger -- 28 Sep 01, 8:45 am

General Forum is maintained by Pi Yee Press