Biased O/U markets?

Pages: [1]
Author Topic: Biased O/U markets?  (Read 1121 times)
VidaBlue
Becoming a Pro
***

Karma: 47
Posts: 236


View Profile

VidaBlue

Biased O/U markets?
« on: November 04, 2019, 10:01:12 AM »

I hang around a lot of mug punters when doing shop betting. What strikes me, is that almost everyone prefers to bet on the "over" market in soccer instead of the "under" market. It kind of makes sense, as it seems more enjoyable waiting for something nice to happen (a goal), rather than celebrating that something does not happen. What I also notice, is that mug punters usually don't care about prices.

Based upon the two statements above, I wonder if it would make sense for a bookmaker to bias the O/U odds slightly in favor of "under" in order to increase the profit margin. Maybe this could be a phenomenon similar to the Favourite-longshot bias https://en.wikipedia.org/wiki/Favourite-longshot_bias.

Do you think this phenomenon could exist in soccer O/U markets?
Logged
tomkruiz
Gaining experience
*

Karma: 10
Posts: 43


View Profile

tomkruiz

Re: Biased O/U markets?
« Reply #1 on: November 04, 2019, 11:57:49 AM »

I wouldn't be surprised. Haven't you mentioned that you've had bad performance while betting on O/U, might it be caused by your inability to remove the juice correctly and thereby not be able to calculate the "real" probability of an outcome? This might call for some data analysis, but if we assume there is a bias on the outcome, you might as well try to apply this assumption when removing the juice and calculating the real probability the next time you try an O/U outcome. Having only two possible outcomes makes it easier. So instead of removing the juice equally from both outcomes, I'd try with 5% steps and for this experiment, I'd allocate some limited resources for the staking plan.

What I mean by 5% steps is, assume the juice is distributed with 55/45 ratio for a fixed number of bets, then by further 5% (60/40) for the same number of bets, then compare the performance. If you feel that the second step has yielded some better performance, try the next step. This might not be the most precise way of finding the answer, mathematicians would probably have a better idea though.
Logged
Wolfie
Pro
****

Karma: 26
Posts: 408



View Profile

Wolfie

Re: Biased O/U markets?
« Reply #2 on: November 04, 2019, 12:27:50 PM »

You can see it in your stats. Are you performing better than expected on unders and lower than expected on overs ?
Logged
VidaBlue
Becoming a Pro
***

Karma: 47
Posts: 236


View Profile

VidaBlue

Re: Biased O/U markets?
« Reply #3 on: November 04, 2019, 12:59:24 PM »

Thank you Wolfie and Tom.

Actually, I intended to introduce the subject as a general discussion of the concept, but since you both mentioned my stats, I will introduce them here and maybe this can contribute to the discussion. These are shop bets:

O/U soccer value bets whole match
Period
2017
2018
2019
2017-2019
Over ROI (# of bets)
    30.0 % (10)
    9.6 % (42)
    -34.3 % (60)
    -10.8 % (112)
Under ROI (# of bets)
    -36.3 % (7)
    -39.3 % (29)
    7.4 % (34)
    -20.2 % (70)

O/U soccer value bets whole match (including either team over or under)
Period
2017
2018
2019
2017-2019
Over ROI (# of bets)
    20.2 % (11)
    12.4 % (117)
    -22.2 % (154)
    -5.7 % (282)
Under ROI (# of bets)
    -12.2 % (19)
    -20.0 % (78)
    -5.5 % (53)
    -14.8 % (150)

The ROIs are real, so of course they depend much upon the conditions at the time of the bet and the staking, but generally both under and over are negative. With a hypothetical flat stake, the general picture would be almost the same.

If I apply 5% steps of juice in favor of the "under" bets, it generally gets worse, since my "under" bets have the worst ROI.

The over bias may exist, but it is impossible to see via my data, as other much larger influences seem to be present. This might be value traps, a phenomenon that I have just recently paid attention to. Maybe the next step is to isolate bets in lower leagues and close to game start. According to another discussion in this forum, these could be good conditions for value traps.

Maybe it still won't be possible to conclude anything based on my own data, because of the limited sample size. I suppose one could match the closing odds for thousands of games in the bigger leagues with their results, for a better approach to a qualified answer to whether the O/U market is biased at some bookmakers. This information is publicly available in vast amounts.
Logged
Wolfie
Pro
****

Karma: 26
Posts: 408



View Profile

Wolfie

Re: Biased O/U markets?
« Reply #4 on: November 04, 2019, 01:04:22 PM »

Maybe is small sample. What software do you use to keep track of data or just excel ?
Logged
VidaBlue
Becoming a Pro
***

Karma: 47
Posts: 236


View Profile

VidaBlue

Re: Biased O/U markets?
« Reply #5 on: November 04, 2019, 01:20:45 PM »

My cellphone counts my steps and captures my betting action on camera, then via OCR relays this information to my remote Apache Hadoop server, including temperature, air humidity and pollution level. Subsequently this is analyzed real time via r's MXNET deep learning library.

No really, for this project I use "just" excel  ;D

Any recommendations?
Logged
Wolfie
Pro
****

Karma: 26
Posts: 408



View Profile

Wolfie

Re: Biased O/U markets?
« Reply #6 on: November 04, 2019, 01:35:31 PM »

Unfortunately no. Just excel is the best answer to any custom made data mining.
Logged
VidaBlue
Becoming a Pro
***

Karma: 47
Posts: 236


View Profile

VidaBlue

Re: Biased O/U markets?
« Reply #7 on: November 04, 2019, 01:42:37 PM »

I agree so much. I think Excel (with VBA) is often being underestimated, even ridiculed, for religious and political reasons. However, if used properly, it can be used for data mining, webscraping, even value bots have been built using this software. But I guess it depends also on what people are used to, no software is perfect, but many overlap and in the hands of an experienced user they can all be used to solve the same task.
Logged
tomkruiz
Gaining experience
*

Karma: 10
Posts: 43


View Profile

tomkruiz

Re: Biased O/U markets?
« Reply #8 on: November 04, 2019, 04:30:05 PM »

I might be wrong, but to my understanding the bias shouldn't apply exclusively to either over or under. When you say you play over or under, is the score margin always fixed? Is it always, say, 2.5, or is it sometimes under 2.5, sometimes over 1.5 etc? I'm asking because this is an important factor in establishing whether a certain outcome, be it over or under, could be overvalued.

Take for example a match between Chelsea and Crystal Palace to be played on saturday. Let's take bet365 odds as an example.
Total --Over --Under --Juice --
0.51.02517.003.44%
1.51.165.006.21%
2.51.532.505.36%
3.52.371.575.89%
4.54.331.225.06%
5.58.001.0834.84%
6.515.001.033.75%

If a favorite-longshot bias exists here, I'd say the bigger portion of the juice is always applied to higher odds, i.e. for 0.5, 1.5 and 2.5 to Under and for 3.5, 4.5, 5.5 and 6.5 to Over. But now I realize I gave you a wrong advice in the previous post. You can't make an assumption that juice is loaded to "real" probabilities in a linear fashion, therefore 5% steps would get you nowhere. Instead, if we assume that your negative ROI implies the existence of the bias (and not value traps that you have somehow kept on hitting), then I'd say the juice distribution ratio depends on the difference between probabilities of two legs of O/U respectively. The bigger the difference between outcomes implied probabilities, the farther the juice distribution is from 50/50.

Does this make sense?
« Last Edit: November 04, 2019, 04:32:02 PM by tomkruiz » Logged
VidaBlue
Becoming a Pro
***

Karma: 47
Posts: 236


View Profile

VidaBlue

Re: Biased O/U markets?
« Reply #9 on: November 04, 2019, 07:45:07 PM »

@tomkruiz

Thank you, your answer makes sense and contains information which is new to me. Somehow I have always assumed that juice was constant throughout the entire line, but clearly it isn't.

So juice seems to diminish in the extreme ranges which does not seem logical - I mean, if the long shot bias exists especially for higher odds, like with under 0.5 (17.00) and over 6.5 (15.00) and juice is already the lowest here, there's not much margin to play with. Maybe these extreme ranges should be disregarded and the study of a possible over/under bias should be restricted to the middle regime where juice is more constant.

I think it is interesting to see that biggest juice is in the O/U 1.5 market and not around the 50/50 outcome. One explanation may be the origin of bet365 odds which are fractional, so juice just varies a bit because it cannot always be the same in a fractional regime (4/25 - 4/1 for O/U 1.5). Another explanation may indeed be connected to the regular punter behavior. They don't do single bets. They all parlay, combining many small odds in order to achieve a lots of small addictive feelings of victory. The over 1.5 bet is the most common one (the next most common is BTS). So maybe it would make sense for a bookmaker to increase juice and perhaps put some bias into this line.

Oh, btw: the score margin that I have played is not fixed, but it has usually been odds within the regime 1.6 to 2.5.
« Last Edit: November 04, 2019, 07:47:30 PM by VidaBlue » Logged
tomkruiz
Gaining experience
*

Karma: 10
Posts: 43


View Profile

tomkruiz

Re: Biased O/U markets?
« Reply #10 on: November 04, 2019, 08:12:12 PM »

Note that these odds were rounded. So for example what seems to be 1.03 vs 15.00 might actually be 1.0275 vs 15.00, which increases the juice to around 4%. The difference between odds is large, and considering they are rounded, we can't make an assumption that the juice is low because of how it seems to be when taking these rounded odds in account. But on the other hand, if there is a favorite-longshot bias applied here, that means the odds of 1.03 are very close to the real implied probability, according to this sportsbook.
Logged
Pages: [1]
Print