Trading What I see !?

[goat68]

.. and Harry was riding a Donkey !

[MobiusGrey]

I've being doing some data analysis and backtesting on UK horses pre-race, using good old horse and jockey form, and modelling of sorts. It's a lot more difficult than dogs...! But unsurprisingly whatever I try based on previous horse & jockey form, ends up breakeven at best, even with the best execution I can find.
I'm not surprised really, it's the obvious thing to do, and the obvious things are never profitable. My dog bot uses a rather different angle on things, and hence why I hope it has legs... Takes 100s of iterations and thoughts to find something others might not have, which is what this game is all about. Pre-race UK horses everyone is at, and have thought of everything, so finding that unique difference that not many other have is going to be VERY VERY HARD!
Hence, everyone is jumping in-play nowadays, as pre-race is too hard, but then that will become the same, and eventually the markets will become a straight line

It's funny you mention this, for a long time I learned to read form, not just horse form but real in depth analysis, monthly subscriptions to various sites to really drill down into run speeds, stride length, sectional times, trainer form, jockey form the whole lot. I was using this to then make a decision based on the race type/track and lay various horses. I never really beat breakeven and coupled with some dodgy financial decisions like using a fib staking system I did worse than breakeven.

The reason that I mention this is out of interest I wondered if I could write a bot that looks at none of this and lays a horse based purely on position in the market pre-off and odds. It's currently out performing me with zero analysis.

[goat68]

I've being doing some data analysis and backtesting on UK horses pre-race, using good old horse and jockey form, and modelling of sorts. It's a lot more difficult than dogs...! But unsurprisingly whatever I try based on previous horse & jockey form, ends up breakeven at best, even with the best execution I can find.
I'm not surprised really, it's the obvious thing to do, and the obvious things are never profitable. My dog bot uses a rather different angle on things, and hence why I hope it has legs... Takes 100s of iterations and thoughts to find something others might not have, which is what this game is all about. Pre-race UK horses everyone is at, and have thought of everything, so finding that unique difference that not many other have is going to be VERY VERY HARD!
Hence, everyone is jumping in-play nowadays, as pre-race is too hard, but then that will become the same, and eventually the markets will become a straight line

It's funny you mention this, for a long time I learned to read form, not just horse form but real in depth analysis, monthly subscriptions to various sites to really drill down into run speeds, stride length, sectional times, trainer form, jockey form the whole lot. I was using this to then make a decision based on the race type/track and lay various horses. I never really beat breakeven and coupled with some dodgy financial decisions like using a fib staking system I did worse than breakeven.

The reason that I mention this is out of interest I wondered if I could write a bot that looks at none of this and lays a horse based purely on position in the market pre-off and odds. It's currently out performing me with zero analysis.

that's the problem every Tom, Dick and Harry (the jockey! still on his donkey) looks at forms establishes their own model totally different from everyone else and thus another varied opinion.... but it's the "net effect" of all these 1000s of models that gives the right answer, but in the long term only...

[ShaunWhite]

that's the problem every Tom, Dick and Harry (the jockey! still on his donkey) looks at forms establishes their own model totally different from everyone else and thus another varied opinion...

That's not the problem that's the chink of light. All these models act at different times at different target prices and therefore if you have a good model you'll find your value price at some point. I thought you'd said you were worried about markets flatlining because everyone would end up with the same perfect model?

[goat68]

that's the problem every Tom, Dick and Harry (the jockey! still on his donkey) looks at forms establishes their own model totally different from everyone else and thus another varied opinion...

I thought you'd said you were worried about markets flatlining because everyone would end up with the same perfect model?

I was taking the... with that statement, as I know humans will always be humans, just look at Parliament !

[goat68]

That's not the problem that's the chink of light. All these models act at different times at different target prices and therefore if you have a good model you'll find your value price at some point.

a good model by definition then, is one that correctly models the average of most of the other models....?

[ShaunWhite]

i'm actually not convinced by horse form models in these UK markets

I love your total self belief. Some guys devote a lifetime to horse form and because you can't making it work after about 20mins trying you discount it. UK horse form models can work, fact, because I know someone with one. But the people with good ones have put in 000s of hours, even just a year full time is about 2,500hrs. The problem however can be finding enough liquid at your price to make it worthwhile.

Btw you're techy, why don't you put every fundamental you can find into ML?

[ShaunWhite]

That's not the problem that's the chink of light. All these models act at different times at different target prices and therefore if you have a good model you'll find your value price at some point.

a good model by definition then, is one that correctly models the average of most of the other models....?

No. In order to get matches you just need to be within the extents of the others' models. 10 models could say 3.0, one could say 5.0 and yours could say 4.8. At some point Mr 5.0 will bet with you.

Also the average/hive mind maybe right on average but models are often about having rare info which is applied to individual markets.

[goat68]

but models are often about having rare info which is applied to individual markets.

So yes, this is what I am thinking, rare/ unique info, as opposed to a model that says if horse raced a class 4 last week and now racing a class 3 then rate it lower priced

[goat68]

i'm actually not convinced by horse form models in these UK markets

I love your total self belief. Some guys devote a lifetime to horse form and because you can't making it work after about 20mins trying you discount it. UK horse form models can work, fact, because I know someone with one. But the people with good ones have put in 000s of hours, even just a year full time is about 2,500hrs. The problem however can be finding enough liquid at your price to make it worthwhile.

Btw you're techy, why don't you put every fundamental you can find into ML?

Undoubtedly takes a huge amount of effort, i probably spent on average only 1.5 hours a day over a year to come up with one dog bot, so thats roughly 500hours, so just to get near your horsey colleague is going to take 5years! You may see why i try and take short cuts...? As otherwise it's not worth it!

[goat68]

So my wonderful 500+hr greyhound bot has lost -£57 today!
Down to £29 profit for the month now, hmmmm

[ShaunWhite]

i probably spent on average only 1.5 hours a day over a year to come up with one dog bot, so thats roughly 500hours, so just to get near your horsey colleague is going to take 5years! You may see why i try and take short cuts...? As otherwise it's not worth it!

He's not horsey at all, he's datary. Try to take heart from the fact you haven't really spent 500hrs on one dog bot, I think you started on them about 2 months ago. I realise choosing a direction is hard but I couldn't understand why you stopped your work on the hft stuff, because that really seemed to suit your skill set and interests.

[ShaunWhite]

So my wonderful 500+hr greyhound bot has lost -£57 today!
Down to £29 profit for the month now, hmmmm

My dog £ is a bit down this year too so far. Do you use any measure other than cash? I'm assuming you rarely get a clean across the board hedge because of the way you close. It can obviously skew your results because it then depends on what wins.

[goat68]

So my wonderful 500+hr greyhound bot has lost -£57 today!
Down to £29 profit for the month now, hmmmm

My dog £ is a bit down this year too so far. Do you use any measure other than cash? I'm assuming you rarely get a clean across the board hedge because of the way you close. It can obviously skew your results because it then depends on what wins.

Correct 2 big losses today due to large SP slip and subsequent wonky hedge...

[goat68]

i probably spent on average only 1.5 hours a day over a year to come up with one dog bot, so thats roughly 500hours, so just to get near your horsey colleague is going to take 5years! You may see why i try and take short cuts...? As otherwise it's not worth it!

He's not horsey at all, he's datary. Try to take heart from the fact you haven't really spent 500hrs on one dog bot, I think you started on them about 2 months ago. I realise choosing a direction is hard but I couldn't understand why you stopped your work on the hft stuff, because that really seemed to suit your skill set and interests.

So the hft stuff the key thing i still dont understand is where the value is...
I know you've gone over this with me a lot, and I apologise (it wasn't a party)
I could offer at prices ticks out and never get matched, i looked at the spike stuff but that never really made sense.
I looked at all sorts of over sold values, vwaps galore, etc
It just never made any value sense, albeit the tech skills were fine, but that's not much good if i don't understand i an looking for....
Where as models and form make sense in my head