[luegofuego]

If Kelly dictates that the shorter the odds, the better it is for your +EG, is there ever a reason to take the favorite on the point spread instead of the moneyline, if you deem both plays to have roughly the same EV?

[u21c3f6]

Assuming that your calculation of your EV (in edge %) is accurate, from an EG (Kelly) point of view you are always better off with the wager that has the higher win %. In fact, up to a point, there are many times where it is still better to wager on the higher win % wager even if that EV edge % is lower than an alternate higher edge % but lower win % wager.

Joe.

[luegofuego]

so why is the common practice usually to take Cleveland -7 for -105 instead of Cleveland -300 or whatever?

[djiddish98]

Typically MLs have a larger vig. So while a 7 spread might be +EV, the ML might be -EV.

[durito]

Typically MLs have a larger vig.

They are usually smaller. Pinny keeps it pretty constant but most other books the vig decrease as the ml's go higher.

[djiddish98]

They are usually smaller. Pinny keeps it pretty constant but most other books the vig decrease as the ml's go higher.

I used the wrong term (thanks for the correction, Durito). I would imagine, though, that the larger spread between the two MLs would make it more difficult to find +EV plays as opposed to the pointspread.

[luegofuego]

why?

[djiddish98]

The higher the ML, the larger the spread you can have without increasing the vig (or even lowering it).

Each additional point you add to a ML has less of an impact in terms of winning percentage than the previous dollar (the further away you get from the +100 line).

Take the no-vig line on the Magic / Wiz game (per Sbrodds for the Greek). It's currently -2350 / +1350, which means the no-vig is ~-1391. If we assume the Greek has the true line here (big if) a 2.5% edge on this game for the Magic ML would be a ML of ~-1013

The point spread is at -110 / -110. Under the same assumptions, a 2.5% edge on this game would be a side at +105.

Which one do you think would be harder to find? The spread that's off by 15 points, or the ML that's off by 1,337 points?

Others, feel free to correct my errors.

[jgilmartin]

so why is the common practice usually to take Cleveland -7 for -105 instead of Cleveland -300 or whatever?

IMO, part of it is psychology. Joe Public likes to bet at odds which are in the general vicinity of 1:1.

[BettingWizard]

ML Favorites are inflated, the higher the point spread


show me one guy that is rich just from betting high ML Favs

[Firefox14]

The higher the ML, the larger the spread you can have without increasing the vig (or even lowering it).

Each additional point you add to a ML has less of an impact in terms of winning percentage than the previous dollar (the further away you get from the +100 line).

Take the no-vig line on the Magic / Wiz game (per Sbrodds for the Greek). It's currently -2350 / +1350, which means the no-vig is ~-1391. If we assume the Greek has the true line here (big if) a 2.5% edge on this game for the Magic ML would be a ML of ~-1013

The point spread is at -110 / -110. Under the same assumptions, a 2.5% edge on this game would be a side at +105.

Which one do you think would be harder to find? The spread that's off by 15 points, or the ML that's off by 1,337 points?

Others, feel free to correct my errors.

Way over my head!

[luegofuego]

The higher the ML, the larger the spread you can have without increasing the vig (or even lowering it).

Each additional point you add to a ML has less of an impact in terms of winning percentage than the previous dollar (the further away you get from the +100 line).

Take the no-vig line on the Magic / Wiz game (per Sbrodds for the Greek). It's currently -2350 / +1350, which means the no-vig is ~-1391. If we assume the Greek has the true line here (big if) a 2.5% edge on this game for the Magic ML would be a ML of ~-1013

The point spread is at -110 / -110. Under the same assumptions, a 2.5% edge on this game would be a side at +105.

Which one do you think would be harder to find? The spread that's off by 15 points, or the ML that's off by 1,337 points?

Others, feel free to correct my errors.

That's just american bias. If you use decimal odds, huge favorites won't be off my 1000 points, they'll be off by 15 or so. The whole premise that MLs are in general less EV seems to be based on... nothing much.

[donjuan]

Limits.

[Data]

if you deem both plays to have roughly the same EV?

They don't. The EV on the ml is much lower. Playing ml instead of a spread is the same as buying points. You pay the full price for the points that worth less to you. Here is an extreme example illustrating the point. Lets say that in the example given by djiddish98 you like Magic so much that you estimate the true line as Orl -50. With the spead like this the points between 1 and 14 are almost worthless and that is the reason ATS play has a much higher EV. Having said that, if you like the fave you will play both, ATS and ml. As the ml will call to place much more money on it comparing to the money placed on ATS bet.

[djiddish98]

Limits.

Wouldn't the limit partially be a function of the line? IE limits are lower for +260 than for -320.

Does it get to the point where any meaningful edge at -2350 still runs up against limits?

[djiddish98]

That's just american bias. If you use decimal odds, huge favorites won't be off my 1000 points, they'll be off by 15 or so. The whole premise that MLs are in general less EV seems to be based on... nothing much.

It doesn't matter if it's decimal or american, you're going to have a lot of trouble finding any sort of edge on huge ML faves, since the spread between the two is so high.

[MonkeyF0cker]

Exposure and bankroll. Suppose you have a -1200 ML with a 3% theoretical edge. If you're betting full Kelly, it calls for you to put 36% of your bankroll in play versus 3.3% against a -110 spread. Not only are you risking a huge portion of your bankroll on one play, but you need to recalc your bankroll minus that huge chunk for any subsequent wagers prior to the completion of the event. That will not play into some people's utility, and also effects your EG on other plays fairly substantially especially on a day like an NCAAF or NCAAB Saturday. Even if you're betting some fraction of Kelly, it is significant.

[MonkeyF0cker]

They don't. The EV on the ml is much lower. Playing ml instead of a spread is the same as buying points. You pay the full price for the points that worth less to you. Here is an extreme example illustrating the point. Lets say that in the example given by djiddish98 you like Magic so much that you estimate the true line as Orl -50. With the spead like this the points between 1 and 14 are almost worthless and that is the reason ATS play has a much higher EV. Having said that, if you like the fave you will play both, ATS and ml. As the ml will call to place much more money on it comparing to the money placed on ATS bet.

This as well.

[TomG]

When you win laying huge odds on the favorite, it's just peanuts. Also upsets happen more than you think. Then when you lose it really hurts. Lay the points instead. That's what they are there for. If you must bet the ML, do it in a parlay to help sweeten the payout a little more with a very safe bet.

[subs]

thank you Data and well explained sir.

a few points your way

[Peregrine Stoop]

ML Favorites are inflated, the higher the point spread show me one guy that is rich just from betting high ML Favs

if a player is wise enough to earn money from ML faves, he's also wise enough to not be "just" betting those

[luegofuego]

Exposure and bankroll. Suppose you have a -1200 ML with a 3% theoretical edge. If you're betting full Kelly, it calls for you to put 36% of your bankroll in play versus 3.3% against a -110 spread. Not only are you risking a huge portion of your bankroll on one play, but you need to recalc your bankroll minus that huge chunk for any subsequent wagers prior to the completion of the event. That will not play into some people's utility, and also effects your EG on other plays fairly substantially especially on a day like an NCAAF or NCAAB Saturday. Even if you're betting some fraction of Kelly, it is significant.

I guess I'm missing something here, but if you're okay with full kelly, I can't see why you would mind having 36% of your roll in play on a 3.3% edge. I mean it would be nerve wrecking, but is there any theoretical flaws to it? Why think of it as "you need to risk a huge part of your bankroll on a play", instead of "you GET to risk a huge bankroll of your bankroll MATHEMATICALLY CORRECTLY"?

[Arilou]

Most people bet the spread because that's what they are trained to do and because that is what they know how to calculate/cap/etc, not because it is better or worse than the ML. For them that is how betting works. It also helps that many players think of "-3 -110" as 3, and think of "-160" as -160, not -147, so they aren't making the correct comparison. At -105 that's enough to make MLs feel bad to you and at -110 it's fatal. Also, a lot of players instinctively worry that the ML won't be "fair" compared to the spread and don't put in the statistical work to know what the right conversion is and/or don't trust a central source like Pinnacle or Matchbook to translate for them.

If you did know all of this, and of course to even approximate Kelly you need to have a conversion written down, then it's a strategic question that varies case by case. There are some spots and sports where it makes sense to bet the favorite on the ML and the dog on the spread while in others the opposite is the case plus there are cases where the lines don't follow their normal patterns for whatever reason and it's up to you to decide whether or not there's a reason for that.

The higher vig heuristic for picking a method is a good one for most people. Don't get involved in lines that are effectively -108 when -105 is waiting for you unless you have a reason. Percent theoretical hold for the bookmaker is not a good measuring stick on the margin (if you think that -37.5 -105 / +37.5 -105 is harder to beat than -1000000 / +10000 then you haven't thought it through) but it's not a bad one for numbers below -500 or so.

[Arilou]

Data, be careful with extreme examples like a theoretical line of -50. First off, if you're giving the game a theoretical line of -50 your model isn't giving you realistic theoretical lines and you need to stop taking its predictions so seriously. Yes, of course if you knew they were actually that good you would look for a bastardized adjusted line like -30 +1000 or whatever you could get, but you know better. In closer cases, it's not a safe assumption to presume that the points aren't worth buying. I'd ask you this: If a team becomes stronger, of the games which it does not cover -10.5, does it win a higher or lower percentage of those games?

[luegofuego]

Most people bet the spread because that's what they are trained to do and because that is what they know how to calculate/cap/etc, not because it is better or worse than the ML. For them that is how betting works. It also helps that many players think of "-3 -110" as 3, and think of "-160" as -160, not -147, so they aren't making the correct comparison. At -105 that's enough to make MLs feel bad to you and at -110 it's fatal. Also, a lot of players instinctively worry that the ML won't be "fair" compared to the spread and don't put in the statistical work to know what the right conversion is and/or don't trust a central source like Pinnacle or Matchbook to translate for them.

If you did know all of this, and of course to even approximate Kelly you need to have a conversion written down, then it's a strategic question that varies case by case. There are some spots and sports where it makes sense to bet the favorite on the ML and the dog on the spread while in others the opposite is the case plus there are cases where the lines don't follow their normal patterns for whatever reason and it's up to you to decide whether or not there's a reason for that.

The higher vig heuristic for picking a method is a good one for most people. Don't get involved in lines that are effectively -108 when -105 is waiting for you unless you have a reason. Percent theoretical hold for the bookmaker is not a good measuring stick on the margin (if you think that -37.5 -105 / +37.5 -105 is harder to beat than -1000000 / +10000 then you haven't thought it through) but it's not a bad one for numbers below -500 or so.

yes, a recent example i came across was when I deliberated Bayern Munich -1.5 for roughly evens or Bayern Munich straight up for 1.32. I mean I don't really know if one holds higher EV than the other, but if one does it's not by much. It seems that you should kinda be more inclined to take them straight up at the 1x2 rather than at the handicap as a default, according to Kelly, while always taking the handicap when betting underdogs...

[That Foreign Guy]

I guess I'm missing something here, but if you're okay with full kelly, I can't see why you would mind having 36% of your roll in play on a 3.3% edge. I mean it would be nerve wrecking, but is there any theoretical flaws to it? Why think of it as "you need to risk a huge part of your bankroll on a play", instead of "you GET to risk a huge bankroll of your bankroll MATHEMATICALLY CORRECTLY"?

There's not a mathematical flaw if you're accurate with your edge estimation. The problem with Kelly, especially on big money line favourites is it magnifies any error in your estimated edge.

As the other poster says it also causes liquidity issues. I don't think I physically could get 36% of my roll down on an event tomorrow (possibly by begging several friends for short term loans). It also complicates bet sizing of any concurrent events you are wanting to bet.

The other reason people don't like betting so much is that human brains are biologically prone to loss aversion. There was a decent discussion of this in the monkey experiment thread a few months ago but in addition to that general bias we hate the idea of catastrophic losses. Our imagination screws with us basically and because we can visualise how much it would suck we sub-consciously assign it a greater probability and therefore feel really uneasy.

[luegofuego]

Alright, cool. But if we're not talking about a spot where Kelly advices you to bet 36% on a -1200 shot, but an ordinary NHL game or something. There you should really be more inclined to take the -300 faves on the ML for 3u instead of one unit at -1.5 for evens or whatever?

[luegofuego]

Also, what's the consensus amount of points to award for a post? :S

[Data]

Data, be careful with extreme examples like a theoretical line of -50. First off, if you're giving the game a theoretical line of -50 your model isn't giving you realistic theoretical lines and you need to stop taking its predictions so seriously. Yes, of course if you knew they were actually that good you would look for a bastardized adjusted line like -30 +1000 or whatever you could get, but you know better. In closer cases, it's not a safe assumption to presume that the points aren't worth buying. I'd ask you this: If a team becomes stronger, of the games which it does not cover -10.5, does it win a higher or lower percentage of those games?

I was merely dramatizing the point which is the same for any given numbers. The points are worth buying only if you are line-indefferent and playing market inneficiencies but if you like a side then they are not. The historical results may vary but if you think about MOV distributions this becomes obvious.