Do you adjust your kelly stake here? I'd guess that it's too correlated to go full kelly on both. I've avoided betting on both totals and sides for a while now, just putting a full stake on whichever I deem to have higher EV. Am I leaving money on the table?
-
-
I bet on both but only up to a total of the largest full Kelly stake on either. It seems intuitively correct and is easy to manage.
I do this on 1H and game total too which seem much more correlated than side and total.
For side and total you could probably bet up to (1+(1-correlation)) * Kely * Kelly fraction.Comment -
I guess I am in the minority, as I look at the sides and totals independently, and if both require a full Kelly stake in same game, I go for it. Sure there will be some "double kills", but the long run is all that matters, so why limit yourself when the edges are there?Comment -
Ah, good enough for me! Thanks mate!Comment -
If they are correlated, I adjust my stakes. Most of the time, they will not be correlated.Comment -
peregrine,
If they aren't directly correlated, as in a -30 favorites and a total of 33 or whatever, they're always indirectly correlated, as in if you have your edge misread on the spread for whatever reason, you probably have it misread on the total aswell. Missing an injury or not weighing some random variable heavily enough or whatever.
There seems to be some amount of dissent on this subject tho, so bump I guess.Comment -
You have to crack open a maths textbook. Something I don't do as much as I should and why I can't help much.
Covariance Matrix
Quadratic Optimization
Modern Portfolio Theory
Those the the leads I got while doing detective work on the same problem.Comment -
There are three cases to consider: Either they're correlated substantially and you can parlay, they're correlated and you can't parlay, or they're not correlated substantially.
The no-correlation case is easy. Place your best separately and party on.
The can't-parlay case is also relatively easy because there's a large correlation (otherwise they'd let you parlay) and you'll need to scale back your wager size.
The can-parlay case is interesting. If you believe that the wagers are sufficiently correlated to impact your betting size, you can and should instead parlay the two wagers! If the correlation isn't enough to prompt you to do that, AND you still want to place both wagers, there's no reason to worry about the bet size.
Note that it is a highly under-used parlay to do an in-game parlay where you like both bets separately and they are only slightly correlated. All things being equal, if you dislike a home team pitcher enough to bet the over and the away team in baseball, for example, the only reason not to parlay them (assuming you can) is if you're running into wagering limit issues.
Note on ProphetOfProfit's post: I'm a math guy myself and I think your decision not to crack the books on this is wise. In general the precise math on things like this will only prevent very small mistakes and soak up time better spent elsewhere.Last edited by Arilou; 05-06-11, 04:06 PM.Comment -
From a cappers point of view, when is a side and a total not correlated at all? In eu football, I can't think of a single case off the top of my head.
edit: Maybe if the expected supremacy is exactly 0.00 and the two teams have exactly the same playing styles. But even in that case, betting both at full kelly should be overbetting since if you're overestimating your edge on one bet, you're almost assuredly overestimating it on the other one aswell...Last edited by luegofuego; 05-08-11, 02:04 AM.Comment -
I just bet half my kelly stake on each.Comment -
In games with small spread and big totals,
Some examples would be NFL games with spread of 2.5 or less, total of 48.5 or more, in NBA games with spread of 3 or less, total of 205 or more.
Those are just two examples where correlation is close to 0.
The lower the spread/total ratio, the lower the correlation.Comment -
IMO, there is so little correlation in these 2 bets that if both bets are indepently worth a full kelly bet than I would bet them as such. Especially pregame NBA where if the fav wins the Over is still never more than 52% in a vacuum and if a dog wins the Under is never more than 52% in a vacuum. Now maybe college football would be different. Ex: Texas is a 24 pt fav in the 1H vs Tulane and the 1H total is 27 (extreme I know but you see a few of these a year). If you thought both Tex -24 and O27 were good bets I would bet them both for maybe 60% of a normal sized bet in that spot being that the correlation in this example is obviously extreme.Comment -
I like turtles.Comment
