[-Dominator-]

While it has been proven that this is a fine place of experts whose astonishing talent and vision for the sport would never even allow them to make any predictions wrong and who would never misuse the words "overrated" or "underrated" in an inflatonary way, the same thing hasn't been done with the way more unreliable bookies. As a result of that I decided to come up with this experimental thread during the boring off-season to evaluate the bookies odds for the current top 10 over the course of the next year, dependant on my own motivation of course.

To determine how well the bookies rate those players, I will give every player a score based on his performance in relation to his odds starting with 0 at the beginning of the season. For a win I will substract the opponents winning percentage, for a loss I will add a players own winning percentage:

Player wins: -p (opponent)
Player loses: +p (player)

Example: Player A vs. Player B, Player A has a winning percentage of 80% and wins.
Player A will then get a -0.2 to his current score while Player B receives a +0.2.

This means that a minus score stands for underrated and a plus score for overrated.


I will take those percentages from the bookies odds displayed on flashscore.com (probably the bet365 ones, but am up for better suggestions). Just for the record, I will use the following formula to get the winning percentage out of the bookies odds. Assume that will eliminate the margin the bookies take for betting:

p (player) = 1 / [ odds (player) / odds (opponent) +1 ]


All in all, for all that are interested in this, what do you expect to happen? Feel free to leave any suggestions, maybe some ideas how to turn this into a funny game (till the next season starts ofc.) or ask me whatever you like about this.

Reminder: This is not a thread to discuss your specific bets. Want to do this, please visit the betting thread:

Betting Thread

Maybe use this for some nice odds and just giving an outlet to write it down... Basha was a 2.55 against Struff, I'm glad I took that on... but I took it in a multi, so now I need Thiem to come through.

www.menstennisforums.com

 

[-Dominator-]

Overview of the Top 10 players I will look at:

1. Nadal: 0
2. Djokovic: 0
3. Federer: 0
4. Thiem: 0
5. Medvedev: 0
6. Tsitsipas: 0
7. Zverev: 0
8. Berrettini: 0
9. Bautista Agut: 0
10. Monfils: 0

Hopefully mods can give me the permission to permanently edit this post so I can update this list whenever I want.

 

[-Dominator-]

Reserved #2.

 

[Zverevdadiii]

Very much looking forward to this.

Nice.

I guess you might revise your opinion of Pospisil while you're at it

 

[Virtualmorpher]

@-Dominator-: how are you going to handle players that get injured and lose as a result?

 

[KingOfClay]

"Betfair exchange" or "Pinnacle" odds would be a much better instead of Bet365. They offer better odds with very low bookmaker margins.

 

[TheLastJedi]

And the purpose is to find out how many Fedfans are running the bookies?

 

[-Dominator-]

@-Dominator-: how are you going to handle players that get injured and lose as a result?

Then it counts as a loss. In case of a withdrawal or retirement I'll just drop the match.

"Betfair exchange" or "Pinnacle" odds would be a much better instead of Bet365. They offer better odds with very low bookmaker margins.

Thought about using "Betfair" first, but the bookmaker margin is already eliminated with the way I use the odds to calculate the winning % (p1+p2 is 1) and I think the bet365 ones are more accessible on flashscore (?).
It's all about he ratio between the odds of both players (I assume the margin is the same for both), that's why I also used the formula in the OP to calculate the % and not just plainly p=1/odd for both players.

Still want to stay consistent with the bookmaker because there are also (slight) differences between the bookies in how they evaluate matches.

 

[Henrik]

I voted "Bookies are absolutely spot on", because that was the option that fitted my belief the best. Of course, this does not mean that the bookies will come out very close to zero for each player. Basically there are two effects that can (and will) make them deviate from perfection.

The first is the genuine randomness aspect. If we have a true dice, we know that the probability for each number to come up is 1/6. Making the reasonable prediction that number 6 will come up twenty times if we roll the dice 120 times is not bound to come true, though. Rather the opposite, actually, as the chance for this to happen is only 0.0973 (assuming a fair dice). All we can do with a fair bit of confidence is to predict an outcome of sixes somewhere in the interval 13-27. Translated to this test, this means we should allow for a fair bit of deviation from zero before we have evidence enough to declare the bookies incompetent.

The second is the systematic effect of bias from the bookies. Some players are likely to develop in one direction or another during the off-season (a young player may develop, an old one might struggle with injuries, or whatever), and may perform above or below expectations. Bookies might see this after a while, but could be systematically overrating or underrating a certain player for a fair bit of time before it is obvious to everyone what has happened. I guess it is here bookies might differ from each other - some catch up to a new situation quickly and others take a bit more time. Since they are all affected by the behavior of the general gambling population, the difference between individual bookies might not be that big, though.

A problem here is to separate random effects from systematic effects. Genuine bias on the part of the bookies is really interesting (for one thing, it implies a possibility to earn money), whereas the normal random fluctuations are kind of uninteresting (unless you are a mathematician or statistician). However, we can be certain on probabilistic grounds, that some of the top ten players will look underrated or overrated, even if the bookies have evaluated the odds perfectly. The situation improves if the test goes on for a longer time, though, as the relative amplitude of random fluctuations will diminish. Thus I would like to encourage Dominator to go on with this interesting test for more than a season, as the chances of drawing interesting conclusions will improve a lot over time.

 

[ZellKFF]

The second is the systematic effect of bias from the bookies. Some players are likely to develop in one direction or another during the off-season (a young player may develop, an old one might struggle with injuries, or whatever), and may perform above or below expectations. Bookies might see this after a while, but could be systematically overrating or underrating a certain player for a fair bit of time before it is obvious to everyone what has happened. I guess it is here bookies might differ from each other - some catch up to a new situation quickly and others take a bit more time.

FAA during his Challenger breakthrough was an absolute gold mine for a couple months

 

[-Dominator-]

Let me just publish the current standings (right) after the AO (will edit the post above as well). Already put into order, from most to least overrated/underrated.

1. Tsitsipas: 1,48
2. Nadal: 1,09
3. Berrettini: 0,65
4. A. Zverev: : 0,15
5. Medvedev: 0
6. Federer: -0,34
7. Monfils: -0,65
8. Thiem: -0,78
9. Goffin: -1,12
10. Djokovic: -1,98

 

[Henrik]

Let me just publish the current standings (right) after the AO (will edit the post above as well). Already put into order, from most to least overrated/underrated.

1. Tsitsipas: 1,48
2. Nadal: 1,09
3. Berrettini: 0,65
4. A. Zverev: : 0,15
5. Medvedev: 0
6. Federer: -0,34
7. Monfils: -0,65
8. Thiem: -0,78
9. Goffin: -1,12
10. Djokovic: -1,98

Thanks for the update. To have a better understanding of the measure, it is good to see how it gradually changes. That Djokovic is the most underrated is natural, since he has won every match up to now. For the very best players, probably the fluctuations in the rating will prove to be quite large, at least early on.

Funny that Medvedev got a zero-rating. Maybe a little counter-intuitive that Zverev turned out to be over-rated, in spite of his slam breakthrough, but, of course, since all matches have equal weight, it takes some time to compensate for the three losses in ATP cup.

 

[-Dominator-]

Thanks for the update. To have a better understanding of the measure, it is good to see how it gradually changes. That Djokovic is the most underrated is natural, since he has won every match up to now. For the very best players, probably the fluctuations in the rating will prove to be quite large, at least early on.

Yeah, that's part of the reason I almost always update my sheet on a daily basis. Thought about updating it here every couple of months, but I guess it is better to do it every week (as long as I am allowed to). Will try my best.

Funny that Medvedev got a zero-rating. Maybe a little counter-intuitive that Zverev turned out to be over-rated, in spite of his slam breakthrough, but, of course, since all matches have equal weight, it takes some time to compensate for the three losses in ATP cup.

Weighting matches is actually something I haven't thought about yet, but similar to using different mechanisms to determine the score (like squaring the probabilities to value bigger upsets higher), I feel like this is just unnatural and against the way odds are determined (bookies already include the importance of the match in their odds) and used (a 2.0 is a 2.0 regardless of the occasion).

 

[IBM]

Overview of the Top 10 players I will look at:

1. Nadal: 0
2. Djokovic: 0
3. Federer: 0
4. Thiem: 0
5. Medvedev: 0
6. Tsitsipas: 0
7. Zverev: 0
8. Berrettini: 0
9. Bautista Agut: 0
10. Monfils: 0

Hopefully mods can give me the permission to permanently edit this post so I can update this list whenever I want.

In case Steve doesn’t allow you to edit anymore, you can always ask a mod to edit certain stuff. Hope we can find a solution for editing.

 

[JediFed]

Wow, a thread I can comment on. If you're trying to sift out systematic from random error, the only way to do so is to look at the standard deviations of your sample and then evaluate whether you would expect to see a deviation within the standard deviation in your sample size.

For example, your errors are going to be 1/(SQRT (N)) where N is the number of trials in your sample. If your overall deviation > the expected number of standard deviations from the norm, and your sample size is sufficiently large enough to establish a normal distribution, then you've established proof of systematic bias.

I see a couple of problems here. 10 players over 50 matches in a year gives you around 500 matches to analyze. That means your error is going to come in around 4.4 percent or so. I don't know if your sample size is going to be sufficiently large enough to establish any meaningful data, and certainly, you're not going to get a sufficient number of trials to establish bias wrt an individual player.

IOW, this is a fun, but meaningless exercise due to the larger constraints you are up against. You could go to the top 20 which would give you about 1k trials, probably the bare minimum to establish any statistical bias over you sample, though again, you'd have issues with the evaluation of individual players, and also betters using this datat to interpret bias would be very much 'buyer's beware'.

 

[Zverevdadiii]

I'd just like to know, not for future purposes but out of interest, who the best moneymakers are.

Say, if you had bet 100$ on every individual for every match, who would leave you with the biggest profit at the end?

Even if Djokovic wins like 90% of his matches this year (lets say for theoretical purposes), it wouldn't be him.

It'd be someone who came through on a couple 5-6/1 odds more than once.. or something in the middle of that. Someone who continously got 2/1 odds but won most. You get the idea.

I bet in football it would be Juventus Turin at the moment, or Liverpool. After Juve's last loss, it would probably be Liverpool.

Wolverhampton and Lazio would be very close though. Same for Oklahoma and Miami in basketball. They have continously played above expectations.

 

[-Dominator-]

As suggested I will be aiming for weekly updates here, but I've decided to just add another post every week instead of editing OP #2. I won't have to deal with the shitty editing issues here wasting my precious edits and everybody can check the progress in retrospect. Win win for everybody.

As for the list, Monfils clearly "on top" of the list now after cleaning up Montpellier and Rotterdam, while Tsitsipas and Medvedev moved further "down".

Update: 17.02.2020

1. Tsitsipas: 1,93
2. Nadal: 1,09
3. Medvedev: 0,78
4. Berrettini: 0,65
5. A. Zverev: : 0,15
6. Federer: -0,34
7. Goffin: -0,64
8. Thiem: -0,78
9. Djokovic: -1,98
10. Monfils: -3,74

 

[-Dominator-]

Update: 24.02.2020

1. Medvedev: 1,32
2. Nadal: 1,09
3. Tsitsipas: 0,79
4. Berrettini: 0,65
5. A. Zverev: : 0,15
6. Thiem: -0,17
7. Goffin: -0,28
8. Federer: -0,34
9. Djokovic: -1,98
10. Monfils: -3,74

 

[FatchMixer]

As somebody that has worked as a bookmaker and a professional tennis trader for about 10 years, I can tell you that the bookies will never be spot on in the sense that the +/- for players will be around 0 for all of them.

Tennis is unique in that players are able to choose what tournaments they play, with things like surface, form, location and more playing huge roles in the odds and outcomes.

There are also lots of external factors that bookies often don't take into account too well (where I along with others have found weaknesses to exploit) such as
-upcoming tournaments (ie. not caring about a 250 when a M1000 is the next week)
-scheduling (odds are released pretty much immediately once the matchups are set, but if a player played until 10pm and then gets a 12pm match the next day, it can have a significant impact)
-draws (players can only beat who they come up against. You can get a pretty easy draw and win 4 matches even at 1.33 odds, but rather than losing a match as an underdog it goes favourably in your +/- total)

The other aspect is that the goal of the bookmakers is not necessarily to offer the most correct odds. They want to get as equal money as possible for both players, so this can skew the odds, especially in the top 10 where casual bettors are more likely to place wagers. Say Federer comes back from his injury at Wimbledon and has an easy first 4 matches where he doesn't look too good, then goes up against a tough yet lesser known opponent (let's say Khachanov). Now, even though the real odds may be something like 1.8 and 2.0 because Federer is not match fit, his name alone can put it to 1.5 and 2.5 because the average fan will want to have their money on Fed rather than somebody he never heard of.

 

[guitarra]

So Tsitsipas has started to close the gap and earn for his „top dog” status.