Trading What I see !?

[decomez6]

so things just crap again, a couple of good days then it all falls to bits...but you knew that would happen didn't you.... but why should there be any model that works? but then no-one would be profitable? and we know at least a few people on here are

public information has a degree of permiability and
publically available bots and variations of such are a measuremnt yardstick . may is not a bad idea to pick the most famous of them all and be on the other side of the book.
just a thought

[goat68]

finely honed and able to react/respond to moving markets/data. This makes it easier to focus on how to best fit a strategy to a market and adapt accordingly. some things are probably work in progress too, as other markets evolve. His YT vids are well put together and I find it's important to revisit them every few months as the journey develops.

Pre-market manual trading on UK horses is hard, so a form of semi-automation is likely to be a better way overall. ie: Entry point manually, Exit point automated. Some bots can do it all though.

.

viewtopic.php?f=52&t=19933

It can be adapted for Pre-Market trading by adjusting entry times on a steamer and can be used on other markets like football where you know the direction of the trade will go.

adjusting and adapting bots is well , but its a bit of a let down when you follow a certain train of thought and end up `kaputt`.

then :
rewind trying to salvage a bot from absolute failureto point where things were promising .
you then hit a point of no return , where you decide to tweak and parsue a different branch of the ¬oak startegy Tree¬

i have posted two bots and the last one looks promising ,and its quite a huddle to adjust.
may be the bot is bloated and convoluted , and there are better ways to do it.
beginners have a level of tolerance that is driven by what else you do outside trading ,and sometimes you could be carrying ideas in your head but is a task to implement them profitably.

please check this one out and tell me what you think .
goat i can relate .
viewtopic.php?f=37&t=23790

you're spot on, I totally relate to this

[goat68]

Hmm, interesting article: https://juicestorm.com/trading-articles ... ir-trader/
anyway, it just eludes to how difficult "algo" trading is...
i've got my Flumine setup working nicely now, and got a linux VM on a fast Macbook to run backtests quickly on which is helping
so I can iterate strategies, but... I just think i'm thinking too simple, the huge number of bots on horses pre-race makes it very hard to find consistent value I think. I'm working on AUS data now, and unfortunately finding the same!
I can plug and backtest all sorts of fancy "value" ideas using many tech widgets, vwap, mvwap, fair value estimator, all sorts of WOM algo's, all sorts of "matched volume" algos, over the long run at best they're breakeven so far....
I am just wondering if the Betfair algo world is becoming an environment where you invest (cost+losses+time investment) to only return at best the same or less. If you take say 1year to come up with a new algo that works, you invest say £2k in achieving it, but the edge disappears 6months later...back to square1,...could be a vicious cycle...

(there should be enough comments there to provoke some good algo responses/discussions!!)

[goat68]

My best algo's so far (best defined by show's promise in backtest, not necessarily profitable!) have been "random"(sort of) choices, ie.hmm lets see what happens if I average mvwap with my fair price scaled by WOM...
No reasoning...!

[goat68]

My best algo so far just popped out of the slot machine, with a bit of a random reasoning choice of params!

AUS pre-race, last 10days, 750 markets, algo traded 297 markets, 2239 orders:

Screenshot 2021-06-02 at 22.33.46.png

Ok chaps, live with this for the next week, where do you think it will end up...? lay your bets please... (use vertical scale, currently at £90)

[ShaunWhite]

How can you test anything with only a few weeks data? And the huge number of parameters in your calculations are far too complex to ever test thoroughly. The first stage in refining an algo is to establish which parameters are relevant by running maybe 10 tests with very very slight variations of ONE param to find if the results display a normalised distribution. If not then exclude it. If it's relevant then move onto the next. But the overriding problem you've got is being over complicated. Eg if you've got vwap in your calc then that alone must have a dozen parameters. How do you propose to test those?

I did say quite some time ago that you were making life really difficult just looking at UK horses, there just aren't enough to suit your approach.

[alexmr2]

Those results look good. BTW do you know about the NSW turnover charge in Aus?

I don't know much about it except that after a certain point (something like £3k of TURNOVER e.g. 6x £500 stakes in 1 day). The commission is so ridiculous that basically people don't run bots on it or can't go over that threshold because it's -EV

[ShaunWhite]

My best algo so far just popped out of the slot machine, with a bit of a random reasoning choice of params!

AUS pre-race, last 10days, 750 markets, algo traded 297 markets, 2239 orders:
Screenshot 2021-06-02 at 22.33.46.png

Ok chaps, live with this for the next week, where do you think it will end up...? lay your bets please... (use vertical scale, currently at £90)

I wouldn't go live with that, I'd monitor it for a while and spend a few weeks really understanding the significance of each variable.

You mention spending 2k on a strategy that works for 6 months, well that's a reasonable return if it makes 300 a day. And this isn't a job where you work for a bit then sit back, it's a proper job that needs a degree of continual effort. And you don't just throw away a strategy that begins to fail, you use your understanding of your parameters to adjust it.

[ShaunWhite]

Those results look good. BTW do you know about the NSW turnover charge in Aus?

I don't know much about it except that after a certain point (something like £3k of TURNOVER e.g. 6x £500 stakes in 1 day). The commission is so ridiculous that basically people don't run bots on it or can't go over that threshold because it's -EV

TOC only applies to thoroughbred horse racing in NSW. There's lots of aus racing besides that. The higher aus commission rates generally don't help much but at least your discount rate still applies to that even though it's no longer relevant for UK racing.

[goat68]

I wouldn't go live with that, I'd monitor it for a while and spend a few weeks really understanding the significance of each variable.

Where's your sense of adventure Shaun?!

[goat68]

Those results look good. BTW do you know about the NSW turnover charge in Aus?

I don't know much about it except that after a certain point (something like £3k of TURNOVER e.g. 6x £500 stakes in 1 day). The commission is so ridiculous that basically people don't run bots on it or can't go over that threshold because it's -EV

so I am aware of NSW TC.
So just checked orders for today, and across ALL AUS markets today total back bets was £1100 so roughly $2k AUD. I can't easily tell which markets are NSW... but say 1/5th were today, that's still $400 a day or circa $3000 a week, so definitely heading that way which is not good!

[decomez6]

Hmm, interesting article: https://juicestorm.com/trading-articles ... ir-trader/
anyway, it just eludes to how difficult "algo" trading is...

(there should be enough comments there to provoke some good algo responses/discussions!!)

‘Garbage in garbage out ‘

Humans have developed very highly sophisticated hardware to allow lightning fast computation, but the software development is lagging behind . Some programming languages are convoluted with bloated logical steps, which makes them
Hard to debugg . This is bad news for intelligence learning . It’s always good to choose your programming language wisely , otherwise you will be left behind and unfortunately there are many intelligent minds that are stuck with the wrong programming languages . The same people needed to bring the dream to fruition.
It’s the human part of the equation that stops the machines from learning .
It also depends on how you see trading :
1. Art
2. Science
3. Mix. Of both

It’s very hard to quantify human behaviour , and that’s why I think manual trading will always exist , so long as humans are making those ALgos. Not unless an algo is in charge of developing another algo , then the market affeciency is artificial in the making.
Take the case of cross matching , it’s a sole purpose was to introduce fair play and balanced books , but then it is an artificial interference of the natural market forces . This in return is a good hunting ground for a trained eye , to spot the mechanical inconsistencies of the market . Some prices are just rediculous and it’s not hard to spot if you wear the right ‘scanning glasses ‘

Also , where are you in the food chain’
1. Whale
2. Shark
3.tuna
4. Cannon futter

We are all chasing the big algo ! Betfair book % equaliser.

[Black Ladder]

Why do some people make some of the simplest things so complicated.

[goat68]

Why do some people make some of the simplest things so complicated.

great joke!

[goat68]

Those results look good. BTW do you know about the NSW turnover charge in Aus?

I don't know much about it except that after a certain point (something like £3k of TURNOVER e.g. 6x £500 stakes in 1 day). The commission is so ridiculous that basically people don't run bots on it or can't go over that threshold because it's -EV

TOC only applies to thoroughbred horse racing in NSW. There's lots of aus racing besides that. The higher aus commission rates generally don't help much but at least your discount rate still applies to that even though it's no longer relevant for UK racing.

What's defined as a Thoroughbred race?