I have a question regarding tennis betting. Is it possible to predict the chance of each score happening in a tennis match/set, if you know how often the player will win points on own serve?
Lets say player A will get first serve in 65% of time. She will win it 60% of the time and 2nd serve will win 45% of the time.
Players B will get first serve in 65%, 1st serve win 70%, 2nd serve win 50%.
Is there any way through math to predict the different percentages of the possible outcomes of this set/match?
Math in tennis betting
- DoctorEvil
- Gaining experience
- Karma: 1
Post
but formulas in question are pretty damn huge and it would take you something like 10 minutes of typing to calculate the probabilities, so the only feasible way of using them is by programming it in some language.
Re: Math in tennis betting
if you assume points throughout the match are independent, tennis match can be described with a Markov chain. so, yes there is an exact formula for the chance of player A winning a match against player B while knowing only how often each player wins a point on his serve. I'm doing some research on tennis betting and I calculated it for 3 variants: best of 3 with 3rd set only going until the tie-break, best of 5 with 5th set going until the tie-break and best of 5 with 5th set not having a tie-break (like grand slam matches).NedlogViiibes wrote: I have a question regarding tennis betting. Is it possible to predict the chance of each score happening in a tennis match/set, if you know how often the player will win points on own serve?
Lets say player A will get first serve in 65% of time. She will win it 60% of the time and 2nd serve will win 45% of the time.
Players B will get first serve in 65%, 1st serve win 70%, 2nd serve win 50%.
Is there any way through math to predict the different percentages of the possible outcomes of this set/match?
but formulas in question are pretty damn huge and it would take you something like 10 minutes of typing to calculate the probabilities, so the only feasible way of using them is by programming it in some language.
- DoctorEvil
- Gaining experience
- Karma: 1
Post
Re: Math in tennis betting
for the match you posted, player A wins a point on his serve (first or second) 0.65*0.6 + (1-0.65)*0.45 = 0.5475. player B wins a point on his serve 0.65*0.7 + (1-0.65)*0.5 = 0.63.
with these parameters in a best-of-3 match, player B should win 85.85% of the time, odds should be 1.16.
for a best-of-5 match with TB in the last set, player B wins 91.00%, odds 1.099
best-of-5 with no TB in last set: 91.19%, odds 1.097
with these parameters in a best-of-3 match, player B should win 85.85% of the time, odds should be 1.16.
for a best-of-5 match with TB in the last set, player B wins 91.00%, odds 1.099
best-of-5 with no TB in last set: 91.19%, odds 1.097
- maletaja
- Pro
- Karma: -22
Post
Re: Math in tennis betting
So what if u do your own analyses? Bookies know quite well they how should be+they add 10% juice for extra. So how do u beta them?They are not mistake more than 10%
- NedlogViiibes
- Gaining experience
- Karma: 3
Post
Some advanced stuff. Will be interesting to dig deeper!
@Maletaja
Not really sure I understand you. If you think the books are spot on with their lines 100% of the time, then there would be no point betting at all.
Re: Math in tennis betting
Thanks for your answer!DoctorEvil wrote:if you assume points throughout the match are independent, tennis match can be described with a Markov chain. so, yes there is an exact formula for the chance of player A winning a match against player B while knowing only how often each player wins a point on his serve. I'm doing some research on tennis betting and I calculated it for 3 variants: best of 3 with 3rd set only going until the tie-break, best of 5 with 5th set going until the tie-break and best of 5 with 5th set not having a tie-break (like grand slam matches).NedlogViiibes wrote: I have a question regarding tennis betting. Is it possible to predict the chance of each score happening in a tennis match/set, if you know how often the player will win points on own serve?
Lets say player A will get first serve in 65% of time. She will win it 60% of the time and 2nd serve will win 45% of the time.
Players B will get first serve in 65%, 1st serve win 70%, 2nd serve win 50%.
Is there any way through math to predict the different percentages of the possible outcomes of this set/match?
but formulas in question are pretty damn huge and it would take you something like 10 minutes of typing to calculate the probabilities, so the only feasible way of using them is by programming it in some language.
Some advanced stuff. Will be interesting to dig deeper!
@Maletaja
Not really sure I understand you. If you think the books are spot on with their lines 100% of the time, then there would be no point betting at all.
- maletaja
- Pro
- Karma: -22
Post
Re: Math in tennis betting
I mean bookies buy statistic from betradar which is quite accurate. Adding killing spread makes is more difficult to beat.
- NedlogViiibes
- Gaining experience
- Karma: 3
Post
Do I think it is the holy grail in tennis betting? No. But I think it will allow me to make better decisions -> less bad decisions.
Re: Math in tennis betting
I want to beat sharps as this is where the real money is. I would be wasting my time, if the goal was to be able to beat soft bookmakers. You also seem to think that bookmakers do not make mistakes - then none of us would be able to make moneymaletaja wrote: I mean bookies buy statistic from betradar which is quite accurate. Adding killing spread makes is more difficult to beat.
Do I think it is the holy grail in tennis betting? No. But I think it will allow me to make better decisions -> less bad decisions.
- Samael
- Gaining experience
- Karma: 7
Post
Re: Math in tennis betting
First of all to think that sharp bookmakers dont have mistakes is just a myth of course they do,it is up to you are you able to see it.
Bookies would be unbeatable if they are able to set your chances at exactly 50-50 every time.Im talking about case where we have odds around 2.00 of course if you have 50% chance of winning on odds 3.50 you have a good bet.
If you open todays markets there are tons of matches with even more side markets and yes some of them are wrong,no way every spread of hundreds of them is accurate.
They have heir own tricks and data,and they also use sharp players who help them adjust odds/lines as most of you know it is easier generally making money on early markets.
Now example of very sharp line today is Karlovic-Muller match.Even average guy knows that we are set to see tie breaks here with logical bet over games.Pinnacle at the moment has that set at over 26.0 1.67 meaning you will need three sets to land this,and with bad odds=zero value.If every match is this accurate we would be in big problems and thankfully they are not.
I get far from the topic sorry.Regarding stats this can be useful but can be dangerous to follow them blindly,
There isnt just serve hold percentage,first serve % etc - there is more things like break stats(from conversion to saving break points) % percentage of points won on 2nd serve can be crucial as well and it is very hard to come up with accurate data as players change tournaments,surface and their percentages also vary troughout the season.Too small sample size can also trick you badly
One of my bets from yesterday Karlovic-Cilic over 10.5 1st set-this one is very simple,karlovic hold% is brutal on grass he will hold 5 games in most scenarios against cilic on fast grass surface.
Cilic is also very good on serve and with Karlovic proven to be too slow on the break again likely that we will see long 1st set (at least 5 games both) as it happened in the end.
There should be a lot of stats to go with this,im bad at math so i dont do that and i dont have them to show
Some people are able to come up with models,but i think you have to know the game as well to be able to distinct false data from something that could give you good bet.
Bookies would be unbeatable if they are able to set your chances at exactly 50-50 every time.Im talking about case where we have odds around 2.00 of course if you have 50% chance of winning on odds 3.50 you have a good bet.
If you open todays markets there are tons of matches with even more side markets and yes some of them are wrong,no way every spread of hundreds of them is accurate.
They have heir own tricks and data,and they also use sharp players who help them adjust odds/lines as most of you know it is easier generally making money on early markets.
Now example of very sharp line today is Karlovic-Muller match.Even average guy knows that we are set to see tie breaks here with logical bet over games.Pinnacle at the moment has that set at over 26.0 1.67 meaning you will need three sets to land this,and with bad odds=zero value.If every match is this accurate we would be in big problems and thankfully they are not.
I get far from the topic sorry.Regarding stats this can be useful but can be dangerous to follow them blindly,
There isnt just serve hold percentage,first serve % etc - there is more things like break stats(from conversion to saving break points) % percentage of points won on 2nd serve can be crucial as well and it is very hard to come up with accurate data as players change tournaments,surface and their percentages also vary troughout the season.Too small sample size can also trick you badly
One of my bets from yesterday Karlovic-Cilic over 10.5 1st set-this one is very simple,karlovic hold% is brutal on grass he will hold 5 games in most scenarios against cilic on fast grass surface.
Cilic is also very good on serve and with Karlovic proven to be too slow on the break again likely that we will see long 1st set (at least 5 games both) as it happened in the end.
There should be a lot of stats to go with this,im bad at math so i dont do that and i dont have them to show
Some people are able to come up with models,but i think you have to know the game as well to be able to distinct false data from something that could give you good bet.
- NedlogViiibes
- Gaining experience
- Karma: 3
Post
Re: Math in tennis betting
Thanks for your input Samael.
"Now example of very sharp line today is Karlovic-Muller match.Even average guy knows that we are set to see tie breaks here with logical bet over games.Pinnacle at the moment has that set at over 26.0 1.67 meaning you will need three sets to land this,and with bad odds=zero value.If every match is this accurate we would be in big problems and thankfully they are not."
1 hour before the match you could get 1.74 on that over 26 bet (it ended in 1.66). If we assume 1.66 was the correct odds, then we would win approximately 2.35% in the long run from a bet on 1.74. I don't know what the limit was at the time, but 1 hour before the match I suspect it to be rather high.
"I get far from the topic sorry.Regarding stats this can be useful but can be dangerous to follow them blindly,
There isnt just serve hold percentage,first serve % etc - there is more things like break stats(from conversion to saving break points) % percentage of points won on 2nd serve can be crucial as well and it is very hard to come up with accurate data as players change tournaments,surface and their percentages also vary troughout the season.Too small sample size can also trick you badly"
I agree that you should not follow models blindly. But the rest of the above I do not agree with.
A model like this would treat the players like robots, so you would only need 1st serve in%, 1st serve win% and 2nd serve win% to create a model. The problem here is that we are dealing with humans, but I still think it is way better to see the players like robots to get a picture than having no picture at all.
Sample size should not be a problem here either.
I do not think this is the holy grail. But I think I will get a better understanding of how prices are set by the market/sharp books. Kind of like expected goals models in football.
"Now example of very sharp line today is Karlovic-Muller match.Even average guy knows that we are set to see tie breaks here with logical bet over games.Pinnacle at the moment has that set at over 26.0 1.67 meaning you will need three sets to land this,and with bad odds=zero value.If every match is this accurate we would be in big problems and thankfully they are not."
1 hour before the match you could get 1.74 on that over 26 bet (it ended in 1.66). If we assume 1.66 was the correct odds, then we would win approximately 2.35% in the long run from a bet on 1.74. I don't know what the limit was at the time, but 1 hour before the match I suspect it to be rather high.
"I get far from the topic sorry.Regarding stats this can be useful but can be dangerous to follow them blindly,
There isnt just serve hold percentage,first serve % etc - there is more things like break stats(from conversion to saving break points) % percentage of points won on 2nd serve can be crucial as well and it is very hard to come up with accurate data as players change tournaments,surface and their percentages also vary troughout the season.Too small sample size can also trick you badly"
I agree that you should not follow models blindly. But the rest of the above I do not agree with.
A model like this would treat the players like robots, so you would only need 1st serve in%, 1st serve win% and 2nd serve win% to create a model. The problem here is that we are dealing with humans, but I still think it is way better to see the players like robots to get a picture than having no picture at all.
Sample size should not be a problem here either.
I do not think this is the holy grail. But I think I will get a better understanding of how prices are set by the market/sharp books. Kind of like expected goals models in football.
- Samael
- Gaining experience
- Karma: 7
Post
Re: Math in tennis betting
I think this topic is very interesting so i will put few more words,also i would like to encourage others to particitape
Im glad i used example of Karlovic-Muller match, which proven too be truly very sharp line as i said.It ended 7:6 7:6
to quote you Nedlog Viibes":
"1 hour before the match you could get 1.74 on that over 26 bet (it ended in 1.66). If we assume 1.66 was the correct odds, then we would win approximately 2.35% in the long run from a bet on 1.74. I don't know what the limit was at the time, but 1 hour before the match I suspect it to be rather high."
From my perspective it had no point point taking 26.0. 25,5 at about 1,70 odds at least would make sense to me as last value line, as it was decent chance we see two tie breaks.
This is example of good job by bookmaker when they made us play 26.0 (or more) which means three sets for win.
I think that chances for three sets here werent that good compared to possibility of ending in straight sets with result like 7:6 6:4 (it can happen even with these big servers).Why would you risk money on that when potential 7:6 7:6 will just give you void
Regarding treating players like robots it can be true,as they are some players that you more or less know what they will bring and how they breathe on the court.One of players that i have on my bets very often is Vesely who doenst tend to (under)perform outside margins i have in my head for him.
Some opposite examples would be players like Paire,Tomic,Youzhny,Troicki which are proven to be very tricky to bettors.With this players there is motivation,tanking,confidence and lot of else included.They dont have a pattern or have it on very short spans.and i dont think they are easy (sometimes impossible) to represent with model
If you watched match between Tomic and S.Robert recently which should end 2-0 for Tomic you will know what im talking about.
I think there was no possible data or common sense that would tell you, Robert will take one set,except sole statement "tomic will stop playing and give it to him"
Im glad i used example of Karlovic-Muller match, which proven too be truly very sharp line as i said.It ended 7:6 7:6
to quote you Nedlog Viibes":
"1 hour before the match you could get 1.74 on that over 26 bet (it ended in 1.66). If we assume 1.66 was the correct odds, then we would win approximately 2.35% in the long run from a bet on 1.74. I don't know what the limit was at the time, but 1 hour before the match I suspect it to be rather high."
From my perspective it had no point point taking 26.0. 25,5 at about 1,70 odds at least would make sense to me as last value line, as it was decent chance we see two tie breaks.
This is example of good job by bookmaker when they made us play 26.0 (or more) which means three sets for win.
I think that chances for three sets here werent that good compared to possibility of ending in straight sets with result like 7:6 6:4 (it can happen even with these big servers).Why would you risk money on that when potential 7:6 7:6 will just give you void
Regarding treating players like robots it can be true,as they are some players that you more or less know what they will bring and how they breathe on the court.One of players that i have on my bets very often is Vesely who doenst tend to (under)perform outside margins i have in my head for him.
Some opposite examples would be players like Paire,Tomic,Youzhny,Troicki which are proven to be very tricky to bettors.With this players there is motivation,tanking,confidence and lot of else included.They dont have a pattern or have it on very short spans.and i dont think they are easy (sometimes impossible) to represent with model
If you watched match between Tomic and S.Robert recently which should end 2-0 for Tomic you will know what im talking about.
I think there was no possible data or common sense that would tell you, Robert will take one set,except sole statement "tomic will stop playing and give it to him"
- NedlogViiibes
- Gaining experience
- Karma: 3
Post
Under 25.5 games was 2.95 on the closing odds, which means you were sitting with a valuebet of 41.2*2.95 = 121.54.
I haven't had the time to test anything yet, and I did not watch the Tomic match. The word "should" kind of stands out though in your sentence. He might have been the favorite to win in 2 sets, but there obviously was a chance for Robert to take a set as well. A model imo gives you a more "unbiased" look at things.
Re: Math in tennis betting
If 1.70 was the last spot with value for over 25.5, then you would give it 100/1.70 = 58.8% chance of happening. This means there would be 100-58.8=41.2% chance of under 25.5 games.Samael wrote: Im glad i used example of Karlovic-Muller match, which proven too be truly very sharp line as i said.It ended 7:6 7:6
to quote you Nedlog Viibes":
"1 hour before the match you could get 1.74 on that over 26 bet (it ended in 1.66). If we assume 1.66 was the correct odds, then we would win approximately 2.35% in the long run from a bet on 1.74. I don't know what the limit was at the time, but 1 hour before the match I suspect it to be rather high."
From my perspective it had no point point taking 26.0. 25,5 at about 1,70 odds at least would make sense to me as last value line, as it was decent chance we see two tie breaks.
Under 25.5 games was 2.95 on the closing odds, which means you were sitting with a valuebet of 41.2*2.95 = 121.54.
I haven't had the time to test anything yet, and I did not watch the Tomic match. The word "should" kind of stands out though in your sentence. He might have been the favorite to win in 2 sets, but there obviously was a chance for Robert to take a set as well. A model imo gives you a more "unbiased" look at things.
- Samael
- Gaining experience
- Karma: 7
Post
Re: Math in tennis betting
25,5 was last line if you wont bet over games.that is exclusively my opinion someone howewer might think different.
Since Pinnacle did very good job here line can be pumped too much so there can be overreaction and make unders value.
While i wouldnt touch this one,technically if you want to bet under games you need at least 26.0 or give yourself a favour,possibility to settle in straight sets and take 26,5.
I forgot what was the odds but in this particular situation is much more important to get the right line if we determined there is solid possibility of two tie breaks,then go for the bigger odds
Shame im terrible with numbers or math so besides this im useless
Since Pinnacle did very good job here line can be pumped too much so there can be overreaction and make unders value.
While i wouldnt touch this one,technically if you want to bet under games you need at least 26.0 or give yourself a favour,possibility to settle in straight sets and take 26,5.
I forgot what was the odds but in this particular situation is much more important to get the right line if we determined there is solid possibility of two tie breaks,then go for the bigger odds
Shame im terrible with numbers or math so besides this im useless
- maletaja
- Pro
- Karma: -22
Post
Re: Math in tennis betting
I think yes its possible to make some tendencies. For exmaple A Zverev, Kyrgious are very tricky players. Main goal is to bet on solid players. Does market knows it? Dont know, but there are at least chance.
Nadal is very solid player. He could be 5-0 down 2-mf set, but he still fights to win it. Others (90%) think about 3-st set.
I have noticed that value is usually 1-st round matches and 2-round and so on. not finals where everyone eyes are directed Market cannot control 128 players match ups on slams!
Today i watched Kontaveit game. She dominated game and she had bad second set beginning and was 2-0 and 3-2 down with break. Betfair offered 2.3 that she wins second set. . I couldnt believe it and bet. Karlovic and Isner type of players. I tend to think that market is quite sharp on big servers. Mayby only if u chase opening odds to get some bets.
But still im cautios about live betting. If u dont have best sources(legal inside info about players, 0-delay info) you are just guy who feed other sharp professionals pockets
It would be nice to beat some sharp book and tennis is very interesting sport for that.
Nadal is very solid player. He could be 5-0 down 2-mf set, but he still fights to win it. Others (90%) think about 3-st set.
I have noticed that value is usually 1-st round matches and 2-round and so on. not finals where everyone eyes are directed Market cannot control 128 players match ups on slams!
Today i watched Kontaveit game. She dominated game and she had bad second set beginning and was 2-0 and 3-2 down with break. Betfair offered 2.3 that she wins second set. . I couldnt believe it and bet. Karlovic and Isner type of players. I tend to think that market is quite sharp on big servers. Mayby only if u chase opening odds to get some bets.
But still im cautios about live betting. If u dont have best sources(legal inside info about players, 0-delay info) you are just guy who feed other sharp professionals pockets
It would be nice to beat some sharp book and tennis is very interesting sport for that.
- NedlogViiibes
- Gaining experience
- Karma: 3
Post
Re: Math in tennis betting
Thanks for the input maletaja
Good point with Kyrgios/Paire/etc kind of players. They are hard to figure out, as they sometimes lose their heads. You need to take some more things into consideration here -> more variables -> harder for everyone. Is this a bad thing?
For your point with 1st round matches being easier to find value. Well I don't know if this is correct. I don't know how many good "tennis syndicates" there are. But I have an idea, that the lines get effective as soon as these syndicates place their bets (on average). I also think they get a chance to analyse most matches on both tours.
With your Kontaveit example, then I still haven't created any model yet (not going to have any time soon), but it would be an interesting spot to analyse deeper to see, if you really had a value bet. Minus value sometimes wins as well
I actually think that the big servers are the matches, where a model like this would be really helpful. Earlier in this thread I pointed out that odds on over 26 games was 1.74 just one hour before the start of the Muller-Karlovic match. Closing odds was 1.66. Margin here was 2.5%, so if you had bet one hour before gametime (1.74), you were sitting with an EV of 2.35% (if we assume the closing price was the right price). Not bad imo. This was a final between 2 big servers btw. So is it possible to find good spots in these matches as well?
Someone has to be the one getting the money from anything - live/pre/whatever. If you are good enough, why should it not be you?
Good point with Kyrgios/Paire/etc kind of players. They are hard to figure out, as they sometimes lose their heads. You need to take some more things into consideration here -> more variables -> harder for everyone. Is this a bad thing?
For your point with 1st round matches being easier to find value. Well I don't know if this is correct. I don't know how many good "tennis syndicates" there are. But I have an idea, that the lines get effective as soon as these syndicates place their bets (on average). I also think they get a chance to analyse most matches on both tours.
With your Kontaveit example, then I still haven't created any model yet (not going to have any time soon), but it would be an interesting spot to analyse deeper to see, if you really had a value bet. Minus value sometimes wins as well
I actually think that the big servers are the matches, where a model like this would be really helpful. Earlier in this thread I pointed out that odds on over 26 games was 1.74 just one hour before the start of the Muller-Karlovic match. Closing odds was 1.66. Margin here was 2.5%, so if you had bet one hour before gametime (1.74), you were sitting with an EV of 2.35% (if we assume the closing price was the right price). Not bad imo. This was a final between 2 big servers btw. So is it possible to find good spots in these matches as well?
Someone has to be the one getting the money from anything - live/pre/whatever. If you are good enough, why should it not be you?
Last edited by NedlogViiibes on Sun Jun 18, 2017 11:16 pm, edited 1 time in total.