Trading What I see !?

[goat68]

for me value is simply eg.a back bet at 6.0 winning 25% of the time

Ok fair enough.

One approach I'm looking at, is laying the favourite. Wait a minute Fox, that's been done to death/solved, you're wasting your time buddy. The initial stats may prove "they're" right.

Lay all fav's since 2012 @ 2% comms @£100


LayFavs.png


But...there are certain conditions when a favourite is overbet and other attributes where a favourite is at a disadvantage. So that may constitute value when laying them. Of course some results will bite you in the arse, but over time, will pan out.

Thanks Fox. I 100% agree with you here, because this is EXACTLY what my Aus Horse bot does. The reason it's done badly the last month, is because a very high % of fav's have won recently.
Here is exactly what my Aus Horse bot does:
Trigger:
- market volume over 4x the value it was at -7mins. Gives a very rough indication the market is "active"
- 1st fav has over 20% of the total matched volume. Sanity check there's some decent volume for 1st fav.
- 1st fav mid-price < VWAP*0.94. A price action trigger indicating the 1st fav is being backed in at this point
Action=> LAY !!!

That's it! I've stopped it now however, because Aus fav's are winning all the time at the moment and 6% commission is a killer for this...

@SilkRun: Told you I was keeping things simple

[wearthefoxhat]

for me value is simply eg.a back bet at 6.0 winning 25% of the time

Ok fair enough.

One approach I'm looking at, is laying the favourite. Wait a minute Fox, that's been done to death/solved, you're wasting your time buddy. The initial stats may prove "they're" right.

Lay all fav's since 2012 @ 2% comms @£100


LayFavs.png


But...there are certain conditions when a favourite is overbet and other attributes where a favourite is at a disadvantage. So that may constitute value when laying them. Of course some results will bite you in the arse, but over time, will pan out.

Thanks Fox. I 100% agree with you here, because this is EXACTLY what my Aus Horse bot does. The reason it's done badly the last month, is because a very high % of fav's have won recently.
Here is exactly what my Aus Horse bot does:
Trigger:
- market volume over 4x the value it was at -7mins. Gives a very rough indication the market is "active"
- 1st fav has over 20% of the total matched volume. Sanity check there's some decent volume for 1st fav.
- 1st fav mid-price < VWAP*0.94. A price action trigger indicating the 1st fav is being backed in at this point
Action=> LAY !!!

That's it! I've stopped it now however, because Aus fav's are winning all the time at the moment and 6% commission is a killer for this...

@SilkRun: Told you I was keeping things simple

The commision @ 6% is a doozy for sure. Even in the example I gave (UK/IRE), the 2% puts it into the red.

Your example looks okay to be fair. I'd be inclined to factor in something that measures the race to be competitive against the fav. ie; Close proximiity of the 2nd/3rd Fav in relation to the actual Fav, maybe the book% combined in some way.

You might recall the Little Acorns thread a while ago, it still has some value in todays markets.

[andy28]

Problem with Aussie racing is that 7 minutes from the off and you have 500 pound matched. My bots are rubbish but one condition I have is that they do nothing until there is 15K matched, that leaves around 90s. Certain days are better but unless it is a metro meeting (Saturday) or a Wednesday there is very little volume compared to the UK. But I guess if you are value betting that doesn't matter as long as you get matched at your price.

I also notice the trading range is massive compared to UK racing (not that I have traded them much. just going off Peters Video's) and they change quick, I just think they are tough, but that's me as a newbie's take on it

[ShaunWhite]

@goat68 instead of 20% TVol on the fav, have you considered vol share compared to price. Eg 2.0 needs 50% 10/1 needs 10% etc. Ie what bookies would call overbacked or underbacked?

And why just the fav? That logic would/could work equally well on any runner. 'Look at the fav' is a manual traders mindset due to the #eyes=2 limitation they have. The only special thing about the favourite is its name 'the favourite'

[goat68]

@goat68 instead of 20% TVol on the fav, have you considered vol share compared to price. Eg 2.0 needs 50% 10/1 needs 10% etc. Ie what bookies would call overbacked or underbacked?

And why just the fav? That logic would/could work equally well on any runner. 'Look at the fav' is a manual traders mindset due to the #eyes=2 limitation they have. The only special thing about the favourite is its name 'the favourite'

Yes, I've used what I termed Relative volume, so price*matchedVol.

I found from analysis 1st fav was the most correlated, I have run against all runners as well.

[ShaunWhite]

.
I also notice the trading range is massive compared to UK racing and they change quick.

Those sound like advantages rather than disadvantages.

I love weak rubbish markets especially when there's 000s a month. You can keep your tight competitive slow moving infrequent UK horses tbh. Fine if you're staking 4 figs but automation works best doing little and often imo. And if you can do Aus horses then you're well on the way to doing UK dogs, and another 600 opportunities a day open up.

Manual and auto are chalk and cheese, find where manual guys struggle and that's where automation can thrive.

[ShaunWhite]

.
I found from analysis 1st fav was the most correlated,

Overbacking/underbacking is maybe a better gauge than a flat number but tbh it's more an issue when a bookie is trying to balance their books than when every bet is potentially with a different individual on an exchange. A fixed amount like 20% will limit you to certain horses though rather than something you can use across the card.

Only working on the fav is sign the algo needs some work. The fact it only works on one horse would be a red flag for me and wiffs of backfitting. In my tests I don't even know which the favourite is, it's just another asset in an asset class (aka market type) . I'm not just looking to trade the most expensive ones.

[goat68]

.
I found from analysis 1st fav was the most correlated,

Overbacking/underbacking is maybe a better gauge than a flat number but tbh it's more an issue when a bookie is trying to balance their books than when every bet is potentially with a different individual on an exchange. A fixed amount like 20% will limit you to certain horses though rather than something you can use across the card.

Only working on the fav is sign the algo needs some work. The fact it only works on one horse would be a red flag for me and wiffs of backfitting. In my tests I don't even know which the favourite is, it's just another asset in an asset class (aka market type) . I'm not just looking to trade the most expensive ones.

The whole "algo" is above 3 lines

[ShaunWhite]

The whole "algo" is above 3 lines

Two too many.

But any black box you throw the numbers into that outputs a signal in the range from - 1 to +1 is fine if it works!

The key, and the hard part, is the - 1 to +1 result from your formula to indicate side and strength, regardless of the size of the inputs.

[goat68]

Another -£30 today, these bots are slowly draining my bank... don't think i've had an "up" day all week...

[andy28]

Thanks for the advice Shaun, and you are right most Aussie mid week racing would trade about 3-4 times that of a UK Dog race.

[ShaunWhite]

Another -£30 today, these bots are slowly draining my bank... don't think i've had an "up" day all week...

At least you've got a decent run of markets to retest on now, to check the actual is approx what the backtest would say for the same period. You're not that time sensitive or staking big enough to be too bothered by your own presence in the data when it was collected.

All part of gaining confidence your test is actually indicative. Without the reassurance you're within a couple of tenths of a percentage you might as well not have data or the tech because you'll never trust it. Your matched bets history must be pretty big on this one now too and that's well worth digging into, eg pl vs signal strength (chunked and averaged with it being a fairly small sample) to confirm your signal is actually doing something.

I'd put it on paper trading mode while you do that, I don't think you'll learn much more about it by running it longer and if the above all checks out you can re-deploy it. Try not to think of it as losing, try to think of it as having bought information you can't get for free. It's only wasted money if you don't gain something from it or you spend more than was necessary, ie kept going when you already had enough info to do some work on.

[Anbell]

for me value is simply eg.a back bet at 6.0 winning 25% of the time

Ok fair enough.

One approach I'm looking at, is laying the favourite. Wait a minute Fox, that's been done to death/solved, you're wasting your time buddy. The initial stats may prove "they're" right.

Lay all fav's since 2012 @ 2% comms @£100


LayFavs.png


But...there are certain conditions when a favourite is overbet and other attributes where a favourite is at a disadvantage. So that may constitute value when laying them. Of course some results will bite you in the arse, but over time, will pan out.

Thanks Fox. I 100% agree with you here, because this is EXACTLY what my Aus Horse bot does. The reason it's done badly the last month, is because a very high % of fav's have won recently.
Here is exactly what my Aus Horse bot does:
Trigger:
- market volume over 4x the value it was at -7mins. Gives a very rough indication the market is "active"
- 1st fav has over 20% of the total matched volume. Sanity check there's some decent volume for 1st fav.
- 1st fav mid-price < VWAP*0.94. A price action trigger indicating the 1st fav is being backed in at this point
Action=> LAY !!!

That's it! I've stopped it now however, because Aus fav's are winning all the time at the moment and 6% commission is a killer for this...

@SilkRun: Told you I was keeping things simple

What price are you taking?

[goat68]

Ok fair enough.

One approach I'm looking at, is laying the favourite. Wait a minute Fox, that's been done to death/solved, you're wasting your time buddy. The initial stats may prove "they're" right.

Lay all fav's since 2012 @ 2% comms @£100


LayFavs.png


But...there are certain conditions when a favourite is overbet and other attributes where a favourite is at a disadvantage. So that may constitute value when laying them. Of course some results will bite you in the arse, but over time, will pan out.

Thanks Fox. I 100% agree with you here, because this is EXACTLY what my Aus Horse bot does. The reason it's done badly the last month, is because a very high % of fav's have won recently.
Here is exactly what my Aus Horse bot does:
Trigger:
- market volume over 4x the value it was at -7mins. Gives a very rough indication the market is "active"
- 1st fav has over 20% of the total matched volume. Sanity check there's some decent volume for 1st fav.
- 1st fav mid-price < VWAP*0.94. A price action trigger indicating the 1st fav is being backed in at this point
Action=> LAY !!!

That's it! I've stopped it now however, because Aus fav's are winning all the time at the moment and 6% commission is a killer for this...

@SilkRun: Told you I was keeping things simple

What price are you taking?

Reverse best

[Anbell]

When you do your backtesting, do you include the first 2 triggers? They are very weird unless they are there for a very specific reason. And that only leaves your 3rd trigger, which might make sense on its own, but unless it is very tightly related to the other 2, then the bot doesnt make sense (for me, for AU horses)

If the magic is in the 3rd trigger, then why not bet later?