-
AFLW 2024 - Round 8 - Chat, game threads, injury lists, team lineups and more.
Kelly betting live test - is it a madman's goal as some quants around here suggest?
- Thread starter Duritz
- Start date
- Tagged users None
Calculating this requires constructing your own push charts. You can even retrofit a vig free push chart of line value from a sportsbook's alternative line market. Take an average from 2-3 though, because some books are off.
However it is better to do it clean yourself, if you have the match data, it'd take you less than an hour with some programming ability. With none (I did this a few years ago), it takes about half a day in excel.
- Thread starter
- Banned
- #28
You'd need to know the market line in the match though, wouldn't you?
You really need to understand how the Gaussian distribution works.
Many natural physical processes (including football) conform to this distribution. If you have a standard deviation (sigma) and a mean (mu) of the margin of victory, you can construct a simple formula to price any > or < points line.
http://en.wikipedia.org/wiki/Gaussian_distribution
http://en.wikipedia.org/wiki/File:Standard_deviation_diagram.svg
Yes you need to have recorded the historical market line for a game. If you don't have this info, your better bet is to retrofit a vig free version from some books.
- Thread starter
- Banned
- #31
Yeah I get the Gaussian distribution... I just don't calculate it based on standard deviations. Being of racing background I don't believe stdev's for Gaussian distributions work well enough. I assume those links will say you price out to three standard deviations?
- Thread starter
- Banned
- #33
I don't like copying charts and things though, I like to work them out for myself. It's the give a man a fish, teach him to fish thing.
I've discussed this before with my syndicate partners, the old fella in our group somewhat takes your stance. He says reality is a funny thing and that sigmas and averages don't properly conform to what is actually happening.
I tell him if it doesn't conform to the Gaussian distribution, then it's another distribution.
That's not the point of what i'm trying to say. I'm trying to say you can calculate this easily for any points line given the right formula.
You're not copying the charts, your reading the ideas and adapting them for yourself. When I was working on my own, I had nothing but simple demo models of the NFL to play with. But that was enough of a spark to make my own model.
- Thread starter
- Banned
- #36
You can tell your old fella (so to speak) that I'm with him. They don't. And I've proven it in my own racing research... this is why I am reticent when stepping into the world of you quants (so to speak
).OK, I'll re-read.
You can tell your old fella (so to speak) that I'm with him. They don't. And I've proven it in my own racing research... this is why I am reticent when stepping into the world of you quants (so to speak
OK, I'll re-read.
I've done some work for a horse racing professional not too recently ago. I looked at 1.5M runners in UK racing over many years of data. He wanted to find the right conversion approach of win odds to place odds. You have flawed methods like the Harville formula, etc. So I decided to empirically derive an equation through plotting a distribution.
The final distribution essentially looked like something to be derived from a power law. It had a massive left skewed spike, and then an extremely long tail. Funnily enough, the distribution looked almost identical to an article I read on the relationship of the magnitude of earthquakes to the magnitude of aftershocks. Reality works in strange ways.
In the end I created a better place formula, but it still wasn't 'smooth' enough to create a tight 300% place market, we needed to slightly compress/expand to fit it properly where there was either small or large fields.
Sounds like a lognormal.
- Thread starter
- Banned
- #40
Yeah, the place market is an interesting one. Recently, a local ex bookmaker turned professional punter, and a bit of a guru here called Dominic Beirne came up with a pretty good working place model. Bookies and the TAB now use it, I am sure he was paid very handsomely for it!
I have another theory on it, which I reckon is a better one, but that's another story. The one the TAB and bookies use now is pretty good.
This bloke's initials weren't MS by any chance, were they?
- Thread starter
- Banned
- #42
Bets for the games for which final teams are out. Half Kelly, biggest EG bet taken, 10k bankroll.
Syd, line, -18.5, 1000 pinnacle @$2.10
Adel, over 15.5 pts, 500 at 2.35 flem sb
Adel, over 15.5 pts, 650 at 2.30 sb
Coll, over 15.5, 400 at 1.94 luxbetia
Freo, h2h, 1150 at 2.70 tab (taken earlier in the week... has drifted!)
Total - $3700
Syd, line, -18.5, 1000 pinnacle @$2.10
Adel, over 15.5 pts, 500 at 2.35 flem sb
Adel, over 15.5 pts, 650 at 2.30 sb
Coll, over 15.5, 400 at 1.94 luxbetia
Freo, h2h, 1150 at 2.70 tab (taken earlier in the week... has drifted!)
Total - $3700
- Thread starter
- Banned
- #43
hawks line -42.5, $600 at 1.943 pinnacle
gws line +80.5, 1075 at 2.01 pinnacle
gws line +80.5, 200 at 1.962 pinnacle
rich line -28.5 230 at 1.909
this total 2105
Total for round: $6065
gws line +80.5, 1075 at 2.01 pinnacle
gws line +80.5, 200 at 1.962 pinnacle
rich line -28.5 230 at 1.909
this total 2105
Total for round: $6065
No it's a power law distribution.
http://en.wikipedia.org/wiki/Power_law
The drop I had was much more extreme though, I had to fit runners from $1.02 to >$500 (we're talking UK racing here).
Yeah much better looking outlays. I think you're still over betting by not properly adjusting for simultaneous Kelly.
These essays are an excellent reference, Dr Bob's been in the game for a while, his edge is eroding (hasn't been profitable on the NFL for years) but still one of the better cappers out there:
http://www.drbobsports.com/essays.cfm?cat=0
and read about simultaneous Kelly.
- Thread starter
- Banned
- #47
Thanks for that. Remember though,I'm using a static 10k BR, which negates the simultaneous play problem. Betting half Kelly on a static BR figure.
My prices are probably too aggressive... I haven't introduced any 62% cap for eg. If they are too aggressive then half Kelly will help square that up.
Last night's game... Had swans to win by 34. On your model, what price would you have the swans to cover an 18.5 point line, if you had them as winning by 34?
No it doesn't.
That's too far dude. Lines are only very rarely that wrong. A 15-16 point difference vs the spread implies a win % going into mid 60's% in AFL. Either increase your Kelly fraction or re-work your model.
- Thread starter
- Banned
- #49
It does though, doesn't it? Given that I am using $10K as a BR to work out the bets, no matter what the actual BR is, then simultaneous plays are not an issue. They're only an issue if you're overbetting as a result of having many plays in play at once, but that's only the case if you have a reducing or increasing BR. I am sure I am right that simultaneous plays don't matter if you're not adjusting your BR.
Yeah I come up with some pretty aggressive differences to the market, I have to say. Thus far I've just been assuming that I know more than the combined total of the opinions of everyone else out there.
The North vs GWS game... I have it at North by 58 points, the market is 80.5... Aggressive!!
- Thread starter
- Banned
- #50
But how can this be proven?
Edit: Wouldn't the way to prove be to look at the final lines for a couple of years, and how far off the final score they were? How often is a line 16 points off the final margin?
