Tennis Betting with Mean Regression Algorithm

[Smiles Tennis]

Few more picks

[Smiles Tennis]

Few more. It's pulling up lots of faves now, which I don't like betting on as much as dogs. 

 

[Smiles Tennis]

Last few for today. 

 

[Smiles Tennis]

todays, not around much so the lower liquidity ones won't be checked throughout the day. 

[Smiles Tennis]

Few more, hopefully can check later for more matches. 

[Smiles Tennis]

Been mega busy with the business move so not had time to make picks, sorry. Hopefully can get some done again going forward. 

Tomorrows picks. 

[Smiles Tennis]

One more challenger match, and I changed the dates from yesterday when  i made the picks to today, when they're being played. 

[Smiles Tennis]

One more challenger match

[Smiles Tennis]

Couple more Challenger matches, that looks like it for now. I had a bit of a eureka moment last night and it might prove to be better long term. Of course we all know that to win long term you need to bet at value odds, so beating the SP at the sharp bookies or Betfair (as that's pretty much efficient) is the only way. 

I realised that my software pulls in the odds from Pinnacle when the matches are released. I would often check the stats of how the players performs at certain odds but using the current Betfair odds. This could result in inaccurate stats being used. 

Last night I thought what if I check the matches very early, and see how the players perform using the opening price. If I can then beat the SP then long term that should work. 

For example. Bonzi v Brooksby, Bonzi's opening price was 1.82 (55%), when his average opening odds are 1.82, he's won 61% (1.64) of the matches. So using the opening odds I can see how he's performed and if I can get a value price on the results stat of winning like a 1.64 shot then hopefully it's all good. 

I "got on" (paper trading) at 1.75 last night, he's now 1.7, so if he does continue to drop then it's a value bet obviously. 

Molcan was 2.6 last night and is now 2.48, Cressy was 1.58 last night and is now in to 1.48. 

Some have stayed around the same odds, can't see any drifters yet. 

It means checking the matches early, which is also better than me sitting down multiple times in the day and checking the matches before they start. 

Anyway, couple more challenger ones I did this morning. 

[harry_rag]

28 minutes ago, Smiles Tennis said:

For example. Bonzi v Brooksby, Bonzi's opening price was 1.82 (55%), when his average opening odds are 1.82, he's won 61% (1.64) of the matches.

I suppose we’re kindred spirits to an extent as I’m using market prices to try and identify value in the anytime goalscorer market (still in the red at time of posting). In terms of the above do you mean when his average odds are exactly 1.82 or within a certain range? What’s the sample size at 1.82 where he has that 61% win rate and what’s the total sample size for all his matches?

Does he show a similar edge at odds either side of the price point, say 1.78 and 1.86?

I suppose where the questing cynic in me is going with this is that you end up with sample sizes for a particular player at a given price that are too small to be relevant and end up drawing an incorrect conclusion with regard to value bets.

[Smiles Tennis]

7 minutes ago, harry_rag said:

I suppose we’re kindred spirits to an extent as I’m using market prices to try and identify value in the anytime goalscorer market (still in the red at time of posting). In terms of the above do you mean when his average odds are exactly 1.82 or within a certain range? What’s the sample size at 1.82 where he has that 61% win rate and what’s the total sample size for all his matches?

Does he show a similar edge at odds either side of the price point, say 1.78 and 1.86?

I suppose where the questing cynic in me is going with this is that you end up with sample sizes for a particular player at a given price that are too small to be relevant and end up drawing an incorrect conclusion with regard to value bets.

I take a list of matches using 15 ticks each side. So for 1.82 I check when he's started between 1.67 and 1.97. I did extend it to 20 ticks each side in the past and it gave similar results. You usually get 30 matches upwards in the results for a certain odds bracket. Seems a bit small to usually rely on but it's all I can get. And over the years of using this kind of research I have checked thousands and thousands of matches, I don't think I've ever had someone be miles out from the market prediction unless they've played about 11 matches at those odds. 

So Bonzi using 1.67-1.97 is 21-13 (61% wins). 

Using 1.62-2.02 so 20 ticks either side, he's 36-20 (64%)

You have extra wins but 2 of his last 4 wins in the second search were at odds of 1.64 and 1.66, so I went back to using 15 ticks and smaller sample as I hope this will be more accurate even with a smaller sample. Thinking about it all the extra matches in the second search (22 of them) will be where he is 1.62-1.67 or 1.97-2.02 so that's quite an amount too add in when they're at the extremes. 

 

[harry_rag]

Thanks for the explanation. I've had a second read back through the thread now and have a better understanding of where you started off (with the regression idea) and where you've arrived at now. My gut feel is that what you're observing is more likely to be random than have real value in beating the odds over a large number of games but I'd be absolutely delighted if your results prove me wrong and will follow with interest.

Do you factor in the opponent's performance at their price as well, e.g. if both are similarly profitable at their price point it's no bet?

[Smiles Tennis]

1 hour ago, harry_rag said:

Thanks for the explanation. I've had a second read back through the thread now and have a better understanding of where you started off (with the regression idea) and where you've arrived at now. My gut feel is that what you're observing is more likely to be random than have real value in beating the odds over a large number of games but I'd be absolutely delighted if your results prove me wrong and will follow with interest.

Do you factor in the opponent's performance at their price as well, e.g. if both are similarly profitable at their price point it's no bet?

No. I have thought about it in the past but I think it will reduce the number of picks, and also take me twice as long to make selections which was taking long enough as it is in the past. 

I am hoping there's value, time will tell of course. I was thinking just now while I was out running, Pinnacle are the sharpest out there, so the players results should, over the long term, reflect their true odds. 

So a 1.82 opening odds player who wins at the rate of a 1.62, wouldn't you think that 1.62 would be their true odds? and Pinnacle odds, if they're efficient, which they apparently are, should move towards 1.62 as the money comes in and they get more information on the player? Pinnacle can get to the efficient price over time before the match starts, so it's kind of like (wishful thinking here) that by having the Pinnacle opening odds, and the players win % (or true odds), you have the two parts of the recipe that you need to grab value before it starts, thinking that a 1.82 shot who wins at a 1.62 rate *should* move closer to 1.62 the closer the match comes.

Or it's all just random and I will call it a day and start from scratch again lol

[harry_rag]

2 hours ago, Smiles Tennis said:

Or it's all just random and I will call it a day and start from scratch again lol

Sounds like the majority of my systems! 

My head's spinning now (been playing chess) so I'll just see how it goes. I can see the logic I just can't tell if everything makes sense of there's a broken link in there somewhere. Obviously you will make a profit if you can identify prices that are bigger than they should be. I'm just not sure to what extent past performance at a given price point is an accurate guide to future results or if it will help you to identify selections that will shorten with Pinnacle. It would be a miracle if every player's results tallied with the expected odds even if the overall market is accurate. I guess if you do this long enough you will resolve the random v predictive conundrum. 

Good luck and thanks for taking the time to reply.

[Smiles Tennis]

4 minutes ago, harry_rag said:

Sounds like the majority of my systems! 

My head's spinning now (been playing chess) so I'll just see how it goes. I can see the logic I just can't tell if everything makes sense of there's a broken link in there somewhere. Obviously you will make a profit if you can identify prices that are bigger than they should be. I'm just not sure to what extent past performance at a given price point is an accurate guide to future results or if it will help you to identify selections that will shorten with Pinnacle. It would be a miracle if every player's results tallied with the expected odds even if the overall market is accurate. I guess if you do this long enough you will resolve the random v predictive conundrum. 

Good luck and thanks for taking the time to reply.

"It would be a miracle if every player's results tallied with the expected odds"

Miracles happen. Google Susan Boyle if you don't know who she is.

[harry_rag]

1 minute ago, Smiles Tennis said:

Miracles happen. Google Susan Boyle if you don't know who she is.

Not when someone might see my search history!

[Torque]

6 hours ago, Smiles Tennis said:

So Bonzi using 1.67-1.97 is 21-13 (61% wins). 

Using 1.62-2.02 so 20 ticks either side, he's 36-20 (64%)

I'm surprised you're using bands this wide as there's a huge difference in the implied percentage chance of winning at either end of the band. At least it partly explains why you don't ever flag up really short prices with your system, which has always seemed strange to me as the Djokovic's and the Nadal's are always really short and they too will either show a profit over time against their short odds or not.

[Smiles Tennis]

15 minutes ago, Torque said:

I'm surprised you're using bands this wide as there's a huge difference in the implied percentage chance of winning at either end of the band. At least it partly explains why you don't ever flag up really short prices with your system, which has always seemed strange to me as the Djokovic's and the Nadal's are always really short and they too will either show a profit over time against their short odds or not.

Taking a narrower band really reduced the selections. Bonzi for eg, 1.82 price, taking 1.77-1.87 so just 5 ticks each side gives results of 5-2 in his whole career. 

Taking that out to 10 ticks each side gives us 15-8 so still not many. I found that 15 ticks would still give a % around the predicted chance, but also give enough matches to hopefully give a signal worthwhile of using. 

As for short odds, I don't really like backing short odds too much so tend to not really check faves below 1.3. The shorter you get you don't usually get too much movement either from opening time to the start of the match. 

[Torque]

10 minutes ago, Smiles Tennis said:

Taking a narrower band really reduced the selections. Bonzi for eg, 1.82 price, taking 1.77-1.87 so just 5 ticks each side gives results of 5-2 in his whole career. 

Taking that out to 10 ticks each side gives us 15-8 so still not many. I found that 15 ticks would still give a % around the predicted chance, but also give enough matches to hopefully give a signal worthwhile of using. 

As for short odds, I don't really like backing short odds too much so tend to not really check faves below 1.3. The shorter you get you don't usually get too much movement either from opening time to the start of the match. 

I see your point in the first sentence, but I still think the bands you're using are too wide. Maybe if you also incorporated the opposing player into your calculations you could get a worthwhile sample in a narrower odds band.

As for the third sentence, the fact that the odds don't move so much I'd have thought was a positive for what you're trying to do in terms of showing where certain players win more than their odds imply when their odds are at a certain level, particularly as the likes of Nadal and Djokovic are nearly always what some would describe as prohibitively short which would mean you get a decent sample to look at. 

[Smiles Tennis]

34 minutes ago, Torque said:

I see your point in the first sentence, but I still think the bands you're using are too wide. Maybe if you also incorporated the opposing player into your calculations you could get a worthwhile sample in a narrower odds band.

As for the third sentence, the fact that the odds don't move so much I'd have thought was a positive for what you're trying to do in terms of showing where certain players win more than their odds imply when their odds are at a certain level, particularly as the likes of Nadal and Djokovic are nearly always what some would describe as prohibitively short which would mean you get a decent sample to look at. 

That's what I was doing before but from this morning I have tried something different after thinking about it in a slightly different way.  I've explained in posts from this morning. 

[Smiles Tennis]

Tomorrow's picks

[Smiles Tennis]

Lots of selections for today were actually for matches due to play tomorrow. I have swapped a few on the spreadsheet and gone from livescore with selections now, instead of just working my way down the Betfair list. 

Just over 3pts profit today, hopefully this new method for picking selections can work. 

Tomorrows matches. 

 

[Smiles Tennis]

Only a few for tomorrow. Many short price (sub 1.3) matches so I don't check those. Niemeier v Watson might become a selection but need Niemeier to move higher than 1.49. There's a gap in the liquidity at the moment and if she reaches that 1.55 currently there then I would make her a pick. 

Edit: I will try to check Challenger matches tomorrow if I have time, no market there at the moment. 

 

Edited July 2, 2022 by Smiles Tennis

[Smiles Tennis]

No challenger opening odds out yet. Not much on Wimbledon, ATP are all too low odds. Cornet match might become a selection but need the odds to move a bit, basically, if Cornet comes down to about 1.83/1.84 then I will back Tomljanovic. 

I will check Challengers in the morning if I get time. 

[Smiles Tennis]

Tomljanovic becomes a pick now with Cornet coming down in price a bit. 

[Smiles Tennis]

Two picks for tomorrow.