College Bowl Strategy

Every year as the college football season enters the bowl games I participate in a local pool. This year there were 48 players. The scoring for the pool that I am in is pick the winner and the points they win by. If you have the right winner you get the difference between your guess and the actual. So if i pick a team to win by 7 and they win by 10, I get 3 points. Now if the team I pick loses I get the point difference again PLUS 10 points. So if i pick a team to win by 7 and they lose by 10 I get 7+10+10 or 27 points. This is done for all 40 or so bowl games before the first kickoff.

Normally, I just find a sight live Bovada and look at what they are calling the spread to be and go with those numbers, and in past years that usually got me in the top 1/4 of players, but never in the money for the pool. Mainly because out of 40 games there are obviously many upsets. So this year I tried something different. I looked at what the point spread was and found a table that told me what percentage of the time a team would win with that much of a point spread. I then applied a random number table to that so that if, for example, a team had a 70% probability of winning if the random number came up 70 or less I would pick that team. If the random number came up 71-100 I would pick the underdog team to win, but only by 1 point in order to minimize my loss in case the favorite actually won. This year I actually would have won had I followed this "strategy" completely, but I let my feelings get involved and changed one game and finished second, but still in the money. My question is this: Was I just lucky, or is there some mathematical way to determine if the random number overlay actually improved my odds?

Quote: kILIMANJARO70

My question is this: Was I just lucky, or is there some mathematical way to determine if the random number overlay actually improved my odds?

There's no such thing as luck, just Variance.

This is actually really similar to a method that I use for a similar type of thing that I won't discuss here. The main difference is that I put together more cards with the other thing, so you get different sets of underdogs.

You can compute, pretty easily, the probability of every single favorite winning to know that is an extremely unlikely event, so that tells us some dogs are likely to win. I really can't think of a mathematical way to PROVE that what you did should have worked, but I can sort of explain why it did, but why it won't necessarily always work.

WHY IT DID:

1.) Assuming that most players are going to gravitate towards picking the favorite by some amount, (often by even more than what the spread suggests) what is going to happen is that, when one of your dogs does win, that's going to hurt them a lot more than one of your dogs losing helps them. Two reasons:

A.) You're intentionally only picking the dog by one point, whereas they are probably picking the favorites by multiple points. Therefore, when you lose, you're not eating a ton of negative points in the process...unless your dog really gets spanked.

B.) The flip side of A is that, when the other players (almost uniformly) pick a favorite by x amount of points on average, they lose all of those points when the dog wins.

2.) Your way takes into account the probabilistic expectation of a certain team winning, whereas blindly picking the favorite assumes that the expectation is zero.

3.) Your way takes into account the lines for assuming the total you want your team to win by. Some people are simply not going to be doing that. Therefore, you're theoretically picking better win totals when you do pick the favorites.

4.) Variance.

---Obviously, if you end up picking a dog to win by one and they lose by a crap ton, that's going to be very bad for you. Therefore, when you do make an underdog pick, you're essentially counting on that underdog not losing by a ton of points.

Example: Virginia/Navy---Navy was a 2.5 point favorite going into the game, so that makes Virginia what, about 42-45% to win? Okay, so had you picked Virginia in that game, given that Navy covered by 39.5 and beat Virginia by 42...that would have been very bad for you.

Example: If you ended up taking Louisiana Tech, then that would have been very good for you, for basically the same reason.

WHY IT WON'T ALWAYS:

1.) Variance---(See above, that, just in reverse)

2.) Potentially VERY high risk.

-This is kind of where you can control your risk a little bit, but if you're doing that for every single game, it can be an extremely high-risk, high-reward strategy. (Some would argue in favor of that anyway, especially with so many people in the player pool...that probably includes me.)

Okay, so there's Variance after you make your Picks, obviously, but there's also Variance before you make them. The Variance comes from what Picks are you going to make.

One example is FAU and Akron. Now, the spread was 23.5, (and was also covered by that much) but the initial spread gave FAU a roughly 95-96% probability of winning. Now, if you take the RNG and it picks 97, are you really going to pick Akron by one point in that game? I wouldn't.

At some point, I would say, given a line of x, I'm not going to pick this team. I'd probably say I wouldn't pick more than a ten-point dog, (so Auburn would have burned me, but not too badly) and just do what you're suggesting for everything else.

Conclusion

Whether you want to call it Variance or luck, it was that to a large extent, but yours is also a great strategy for a pool of such a high volume of players who are going to be picking mostly favorites. When you do hit a winning dog, it's going to really hurt most of the other players.

But, the Variance comes when you have to land on the right dogs. If you end up picking fourteen to sixteen dogs throughout, (which is probably right about the expected amount, given your method) if you only hit three or four, you'll be completely screwed...obviously.

Finally, since most people are picking the favorites anyway, when a dog you did NOT pick does it, it doesn't really hurt you relative to most of the other players.

Thanks. I read your PM before I read this. I think you see it pretty much as I do. When I entered I told the pool organizer I was trying a new stategy that I call Heaven or Hell. This year it did work out. I did miss the Navy game,,, that was the one I went against the RNG, thinking that they might be down after losing to Army. At any rate, having about 40 games in the pool I think the law of large numbers came into play a little bit as after the Navy game I was at a low point of 22nd place, but still managed to recover by the end of the tourney. Thank you for your detailed explanation!

I wouldn't say, "Law of Large Numbers," because that's not a large number for a sample. I think what happened is simply that enough of your dogs were correct that you were successful. Your dogs could have been wrong. The RNG could have, "Told you," to pick whoever played Navy, in which event you came in second partially because of the RNG as opposed to going against the RNG.

The RNG, "Tells you," to pick Virginia (just looked) 45% of the time as it is...so it is almost just as easy that it would have told you to do that. You really can't look at the thing after the fact and say, "That kept me out of first place," because it could have just as easily been the other way. Besides, depending on how far you finished out, perhaps there's another game you could look at and say, "If the RNG had done..."

top 5
641 avg 16.03
656 avg 16.40
687 avg 17.18
694 avg 17.35
703 avg 17.56