The most profitable bots

[jimibt]

Weird means an angle that no-one talks about.

Although an order flow trader would argue otherwise:
"I have an idea about where price should go, but I want to see someone else go into the trade first"

Just because someone goes into that trade first, it doesn't mean that the direction will be correct. Or that he went into the trade for your same reason. It might well be an outright bet or lay.

In fact, in the context of botting, it's not really relevant whether you pick the direction correctly on the individual trade. It's not even important that you pick the direction correctly more often than not.

It's important, over many trades, if you made more money with winning trades than lost money with losing trades.

divide and conqer - lol

[ShaunWhite]

Weird means an angle that no-one talks about.

Although an order flow trader would argue otherwise:

There's more ways to automate than there are ways trade manually, everything from looking for blue moon events to HFT. The typically manual techniques you're talking about tend to be less popular as they need 'soft' skills.

To paraphrase northbound, it's a balance between being a little bit right a lot of the time, or being a lot right a little bit of the time.

[ruthlessimon]

Just because someone goes into that trade first, it doesn't mean that the direction will be correct.

(bear in mind I'm a longterm swinger)

Although I haven't seen the data yet, I'm convinced I've improved some highly algorithmic biases (which probably should be automated), by reacting to movement, rather than preempting the movement (in the majority of cases). i.e. getting a "Worse entry", paradoxically, improves the expectancy - because I can get better entries - on the original entries that would've taken an initial drawdown

[ruthlessimon]

There's more ways to automate than there are ways trade manually, everything from looking for blue moon events to HFT. The typically manual techniques you're talking about tend to be less popular as they need 'soft' skills.

To paraphrase northbound, it's a balance between being a little bit right a lot of the time, or being a lot right a little bit of the time.

I do agree. but I have to justify why I'm manual - when I probably should be auto

I'd love to understand a bit more, just why the real big manual traders struggle to go auto - if all edges are "simple"

-Because my edges are indeed simple, but they make up for that - by not being that good

[northbound]

I'd love to understand a bit more, just why the real big manual traders struggle to go auto - if all edges are "simple"

My guess would simply be that most people struggle with technology.

Also most people can feel things but are unable to break them down into numbers or logic constructs.

Also perhaps order flow trades are not that simple to automate: one thing is to take snapshots at certain moments in time and build rules around that.

Writing software to interpret real-time flowing data is a whole other thing entirely.

[ShaunWhite]

I'd love to understand a bit more, just why the real big manual traders struggle to go auto - if all edges are "simple"

Manual traders have a lot more info to go on, and a very powerful fuzzy logic engine in their heads.

btw do they struggle to go auto? Or do they just like the buzz of playing the ladder game more than they like looking at code & data?
Most top manual traders I know also like sport, so far the quants I've encountered aren't so keen. Personally if I had to watch sport all day I'd be looking for another job, gimme a screen full of code & numbers and I'm happy as Larry. People tend to excel at what they enjoy I think, no point persuing a method if you don't enjoy it.

[ruthlessimon]

Manual traders have a lot more info to go on, and a very powerful fuzzy logic engine in their heads.

My guess would simply be that most people struggle with technology. Also most people can feel things but are unable to break them down into numbers or logic constructs.

I was just thinking about how I could actually automate my trading, & even though I'd class it as simple, & definitely automatable (on a market by market basis) - the code would have to be incredibly long to encompass the whole day, & forthcoming days - i.e. stuff I was ready for, but didn't occur

So it's the not the fact I can't break it down into a basic logic construct - but scaling it into a big logic construct would be hard.

Whereas the generic "front run the £10k+ bet for 1 tick" would be easy to automate (albeit might eventually require a investment in Dublin, London or a BF employee for those extra milli's)

-Basically, looking at it now, I think it's actually pretty easy to underestimate just how many "what if's" need to be accounted for if going full auto

[ruthlessimon]

-Basically, looking at it now, I think it's actually pretty easy to underestimate just how many "what if's" need to be accounted for if going full auto

Here's actually quite a nice example from today actually

You've got your pure syndicates in @ 2.68 (most profitable simple bots)

But visually I could see the "intent" of that money (i.e. there's nothing @ 2.72) - & I can override my initial plan (which was to back the fav) - because I know this guy still needs £15k filling

& just automating that line ready for all future occurrences would be a total nightmare

[northbound]

Simon, Most probably the way you trade is more suited to manual trading on the ladder rather than automation.

IMHO automation is more suited to a quantitative approach that targets a specific scenario (the less rules/conditions the better).

[PeterLe]

I always remember a comment from PeterLe on here (for those who don't know he's solely an automated trader who has consistently made very good money year after year on in-running horse racing markets)

He once posted if he were to sell his most profitable bot for £20, buyers would ask for there money back as it was so simple

Thanks Dallas
Yes Id still say that comment was as appropriate as ever.
One way for a new trader is to visualise a race in play and then and then take a view as to what the prices should be.
For example, yesterday I was lucky enough to be guest at LIngfield. For the second or third race an injured jockey took us down to the last fence from home (I think it about two furlongs fair distance anyway)...
As the horses took the last jump I had my mobile connected to the VPS and could see the prices live as the horses jumped..
In this particular race the two horses jumped almost together and I saw the back price on one of them drop to about 1.15 (I couldn't see exactly as the prices were refreshing very fast on the API, but I think it was about 1.15, but I was viewing through my beer googles too!)
There was no way 1.15 was value on that horse which incidentally lost too
Having jumped slightly in front of the other, logic says it should have been below 2, but not 1.15?
Something similar will probably happen again tomorrow and the day after and the day after that. There are at least three ways I can think of off the top of my head to take advantage of that..and thats one way to plant a seed of an idea
Regards
Peter

[spreadbetting]

I always remember a comment from PeterLe on here (for those who don't know he's solely an automated trader who has consistently made very good money year after year on in-running horse racing markets)

He once posted if he were to sell his most profitable bot for £20, buyers would ask for there money back as it was so simple

Wouldn't surprise me , so many people overlook the obvious or only listen if you're telling them what they want to hear.

[ShaunWhite]

I was just thinking about how I could actually automate my trading, & even though I'd class it as simple, & definitely automatable (on a market by market basis) - the code would have to be incredibly long to encompass the whole day, & forthcoming days - i.e. stuff I was ready for, but didn't occur

So it's the not the fact I can't break it down into a basic logic construct - but scaling it into a big logic construct would be hard.

Thats because you're not a systems designer

But your take on automation is askew, You don't have one behemoth that deals with everything, you have many micro instances each dealing with specific scenarios. Often more than one will be involved at any given time, and they might be agreeing or taking contrary positions to reinforce or reduce the net effect of the signals involved.

You need to be satisfied with building one tiny part at a time, letting it settle and than adding more elements.

Btw automation isn't just about speed, it's about doing the same thing absolutely unerringly 247 while simultaneously looking at every selection in every market for every event (in multiple sports) It's like having 20 people all doing your job and thankfully they don't need paying, so if they only bring in £10ea you're quids in.

[ruthlessimon]

Thats because you're not a systems designer

Bitta data snooping here, a bit of P&L fudging there, add the odd formula error - & you've got a recipe for a "fun" afternoon

[Trade 500]

I’m interested in what bots are working for some of you, so far I haven’t heard any and am even more curious now.

Personally, I so far invested 1.5 years into trading, collecting unique data, testing ideas, etc. So far, I came up with only a couple of profitable bots. Here's a very remote hint of what they do:

Horses pre-off trade
Based on certain market metrics, it lays a horse X minutes before the off, then closes the trade at scheduled off time. It does around 10 trades per month: some win, some lose. Given a trading stake of £100, it makes a profit of £20-60 every month. Here's the thing: I tested this for 12 months with £2 stakes. It performed really well every month with the exception of Jun-Aug. Only last month, after a year of testing, I was confident enough to raise stakes. Going well so far.

Greyhound value betting
Based on certain market metrics, it backs a greyhound to win the race. It places 200+ bets every month. If you bet £1 per dog, you can expect at the end of the month to be up £10-20. Bear in mind that almost every month at some point this strategy finds itself down £10+ because of losing streaks. What matters is the bottom line after hundreds of bets at the end of the month. Again, this has been profitable every single month since May (when my data recording began) but only in the last couple of months I felt confident enough to raise the stakes.

Thanks northbound, very interesting and I can tell from reading yours and others posts just how much testing is needed to get something that works, and of course how once found it is wise to not share everything as I guess everyone would try it out and then may mess up the strategy a little if hundreds were doing the same thing.

[ShaunWhite]

then may mess up the strategy a little if hundreds were doing the same thing.

If only it were 100s. Many markets are so small that even just one other person trading the same niche can change it's behavior. There are many strategies of course and some are more resilient than others.

It's always been the case though that building a automatic money printing machine will take many many hours of tinkering in your mancave. But if you roll your sleeves up, sharpen your pencil and refuse to be beaten, it's doable. But like most homemade machines, it'll take constant nursing to keep it running.