Forum

[baddydaddy67]

More good contributions. I am making good progress on AUS horses and dogs, UK dogs - Testing strategies with no real trap bias and a reasonable spread of favourites winning; sustainable profit is proving harder.

Beerio, I like 5 horse idea, I do something around laying F3/4 at a reduced price.

One question - for those dutching are you using recovery staking - if so which one? I am using lay X%/Y%.

[Beerio]

More good contributions. I am making good progress on AUS horses and dogs, UK dogs - Testing strategies with no real trap bias and a reasonable spread of favourites winning; sustainable profit is proving harder.

Beerio, I like 5 horse idea, I do something around laying F3/4 at a reduced price.

One question - for those dutching are you using recovery staking - if so which one? I am using lay X%/Y%.

5 Horse races are steady income if laying 3rd and 4th favs, just don't get carried away with the prices. Find the happy middle where you can make 1-3pts per day steady, and if goes terrible then then you only losing 3 or 4 pts that day, but there are way more good days then bad ones.

[Beerio]

I thought on UK dogs with those real tight turning tracks to look at the wider boxes and choose dogs that have raced recently or have finished in top 3 beaten a short distance

no point being a resuming dog drawn the inside
no point not being fit as you need to hit the outside and swoop down

and you need a strong healthy dog in some sort of recent form - I doubt out of form dogs jump well

anyway I think there are filters and sequence you could build on those tight tracks - but I have not done so myself in terms of building something

I haven't gone too deep into the tight turn angle, however there are a few tracks where the favs only win 25% & 27% respectively and im dutching around them, as other tracks will show 30%+ hit rate.

[Beerio]

when it comes to horse racing I have been working on 5 6 7 8 horse fields and that is it myself -- there are way more patterns and in very small fields often fav 1 gets tripped up by the pace or position - the better riders tend to show more smarts.

anyway I got to go out now -- back later

I just autoloaded all sorts of horses and greyhounds for Aus for the overnight sessions. I will run these in sim mode in the times that my actual bots don't run to gather the information to start building a theory. I like the idea you presented with the smaller fields as that's manageable, even tho in big races there's more chances of not F1 winning, but if the horses are 100/1+ then they don't actually have a chance of winning but more hampering a runner or 2 IMO, but I could be proven wrong, so lets see what the data says after a few weeks

[wondra]

Hi

Well I am experimenting with Sequences v Starters.
I had about a 3 weeks worth of stats which I split looking for patterns.
I wasnt able to stop/pause the Strat after 5 losses and restart after a win, so had to try something else. I figure if I am stopping then restarting immediately after 5 losses, as well as restarting after a win, a sequence longer than 5 digits would be redundant.

Anyway, I can up with multiple seq for different size fields. Only been running this for 3 days. On days off now so can do a bit more analysis (see results below)

results 10.07 (1).jpg (100.79 KiB) Viewed 40922 times

results 10.07 (2).jpg (82.96 KiB) Viewed 40922 times

Too early to be definitive however there are early pattern emerging. The simple reason for not experimenting with different seq for all size fields (I ran out of time whilst having to do my day job )
I will analyse the leading contenders to see if I can make an adjustment to get the SR up
I may drop a few of the unproductive seq and try a few other seq.
I also intend seeing if I can link several strats and get them to start after another strat has a certain number of wins, ie,.. Fav 2 strat begin after Fav 1 Strat has 2 wins (or something along those lines)

Anyway, just wanted you to know I am reading and learning from the contributions on this thread.
I will update as I move forward
Thanks everyone for inspiring me

cheers
wondra

[wondra]

thanks for the suggestions...
paralysis by analysis is starting to become an issue

for the past hour I have I have been looking at narrowing down my fields...I have also changes the seq in a couple that arent working.

Im thinking going forward to set up another VPS, so I can have one live betting, the other trialling Strats

Thanks for starting these threads...it is certainly helping me

cheers

[Beerio]

Since the last update, I have combed thru all the GH UK tracks and found a few little nuggets in the results tab, however I have a question how you guys/ladies base your decisions....

What do you find works the best for you? Is it last 30 days worth of results? or overall year by year stats? or combine the them all to find a baseline model and then just go by that? Many thanks in advance.

[Beerio]

I am rapidly losing interest in updating this thread with new stuff

there only seems really 3-4 people who even make an effort besides me -- either no-one can figure it out or no-one cares much about making money

thanks to those who participated so far I appreciate it - I guess we see how it goes bit longer.

Some find it too hard to follow, others find it too difficult to put the research time into building sustainable bots going forward. As much as you want to help people, unless you spoon feed them then they don't wanna know you. As always it's way easier to get things handed to you rather then doing the work required.

On a positive note, after gathering information in the Aus Horse markets, I have built my first bot . It's really early and it won't see many runs daily, but from what I can see it should be +2-5pts or worst draw down for the past 30 days has been -3pts. Overall it was +22pts for the last 30 days. Happy to be stacking up the points.

[Beerio]

No worries, happy travels and keep on improving.

[baddydaddy67]

Dont get bored with the thread - it made me step back and look at the best way to dutch horses and dogs - a mix of F1/F2 or F1-F3 dutching - all done with recovery. I am only dutching this to £1 stakes - mainly to test out strike rate. I made my strategies up using the data I had available through the results tab and did the pain staking checks of every track in UK & AUS - so at the moment running about 120 unique strategies - only one week in. Week one saw a peak profit of £50 only for it to drawndown to a min drawdown of -£8 - now sitting at +£10.

I am using a loss recovery methos - increase stake 20% on losing bet and decrease by 10% on winning bet. Some market strike rates in 80% others 40%+, objective is to continue to tweak the strategies over the whole of July to either:
a) ditch the course / track due o less preditable results
b) Revise the dutching either up or to cover more or less of the field
c) Look at dutching by traps if favourite number dutching is not profitable
d) Revise loss recovery and staking methods to improve profitability
e) Use stop at a profit across strategies to lock in profit on a "Good Run"

The data I have so far doesnt merit posting on this thread yet - maybe in a week or two.

[woweee]

Phew

Having read through all this thread. So much information. I’m so confused. My brain is totally over loaded. So many questions I don’t know where to start.
How can a sequence of winners from the past be used to make a winning system !!!

Could someone explain in layman’s terms ?I
Is the sequence referring to the size of Stake or Favourite/Stall ?

[madjon]

Phew

Having read through all this thread. So much information. I’m so confused. My brain is totally over loaded. So many questions I don’t know where to start.
How can a sequence of winners from the past be used to make a winning system !!!

Could someone explain in layman’s terms ?I
Is the sequence referring to the size of Stake or Favourite/Stall ?

The sequence is favourite position.

[woweee]

Thanks Madjon

How many races would we need to know which fav was the winner ?

Once we have a huge list of results.
Example
3,5,1,6,9,1,5,2,6,4,1,1,4,7,6,3,4,1,6,7,3,1,2,1,5,4,3,1,8,1,2,4,1,2,1,5,3,6,3,1,7,3,2,7,3,1,5,1,6,1,8,3,2,5,1,3,6,3,5

In the above results the sequence 1,2 appears 3 times.
Is this what were looking for ?

[danzak]

This is my example for SQ. See attachment. Good Luck!

[Drizzy12]

DN,

Can I clarify again (as you have mentioned it many times) I never asked you to post YOUR strategy, I asked you to post an EXAMPLE. Because myself, like others in this forum still don't really understand your method, despite trying our best to understand.

With regards to purchasing strategies, I don't think people expect to purchase and then start printing money. I think it's a case of getting some starting inspiration (which many of us struggle with) then tweaking and tweaking some more to tailer it to our needs (that's what I am doing anyway)

Although I don't post (because I have tried and failed to engage) I do enjoy these posts and look forward to your sample strategy you appear to be creating. Can you give us any hints what it is based on?

[echelon]

Hi DN,

A lot's happened since I last looked in on one of your threads.

The SQ ideas in this thread are interesting.

I'm looking at the BFBM results downloaded into Excel and trying to find for example a sequence for Faves 1, 2 and 3 for UK dogs (as in attachment).

The first thing is to look at "view statistics" in the BFBM results screen which gives % occurrence of F1, F2 and F3:

F1 = 30%
F2 = 23%
F3 = 16%

So, if we had a sequence which was 30 + 23 + 16 = 69 steps, then if the sequence was just randomly jumbled 1,2,3's we'd need to have 30 1's, 23 2's and 16 3's ideally as part of the sequence. However, because we reset to the start of the sequence on each winner, we're not going to realistically get anywhere near 69 steps - because it would mean 69 losers in a row.

So, when we create a sequence we'd make it smaller, though in similar proportions to what we've seen in the "View Statistics" - hence these could work well:

F1 - 8 steps
F2 - 6 steps
F3 - 4 steps

From here I think the only clue we can get as to how to structure our sequence is to look at repeating favourites appearing in runs in the results - such as we may see clusters of 1,1 or 1,1,1 in a lot of results, whereas all the 2's and 3's are generally spread out.

Our first step in the sequence must be the favourite we have selected with the highest % frequency from the statistics and the order in which favourites appear in the first steps should be in decreasing order of their %. Hence, F1 in this case is the first step and F2 should appear before F3.

Now - we're in the trial and error phase where we build several strategies in Sim based on the rules above...

Example sequences could be:

1,1,2,1,3,1,3,2,1,1,1,3,2,2,3,1,2,2
1,1,1,1,2,2,2,3,1,1,2,2,3,2,1,3,1,3
1,2,2,2,1,3,1,1,1,3,3,2,1,2,3,1,2,1

Am I on the right track here?

[echelon]

Many thanks for your advice

I prefer to pick up as many races as possible because the variance spreads out more nicely that way - which I think you mentioned in an earlier post.

I like the idea of ignoring any Fave 1 bets below 1.9 because the reward isn't worth the risk.

If I'm going to include 4 and 5 then there's filtering I could add to reduce the amount of losers on these (from what I've seen, longer distances, mean longer running times which means that favourites will be more likely to collide more often allowing less favoured dogs to take the lead). However, including those sort of filters I think would be more suited to a "long shot" strategy - which perhaps concentrates on getting big winners from faves 4,5 or 6 - perhaps with a sequence too?

If I include the distance filter when I have faves 1,2,3,4 and 5 then I suspect that I might be filtering out a lot of profits which could come from the lower faves in short distance races. Though food for thought and something to try out.

You mentioned creating 20 strats in Sim and running them for 1 day. At the end of that day's testing do you recommend ditching all the losers and continuing to test the winners? If so, how long should we be testing the winners for in Sim, before putting them live? A week would cover every day and there are noticeable differences in UK dog results at weekends compared to weekdays - so my guess is that we should do at least 7 days testing.

[wondra]

Hi

Just wanted to say i still following the thread. Work has been busy so tak8ng up a bit of time
Im continuing testing, refining, testing again
I feel im pretty close to going grom Sim mode.
Will update hopefully early next week, working 6 days straight at present.
These latest posts have given me a few new ideas, i have an exercise book with quite a few pages of furure strat project ideas

Thanks
Wondra

[echelon]

also as well I just want to say - your so far thinking about it in terms of absolute

fav 1 this fav 2 that etc

but that is how everyone else thinks - like the guy from betangel in his video

you absolutely need to use filters to defeat the inherent bias against winning by staking fav 1 .... what are the conditions that allow fav 1 to win @70% instead of 35% -- this is key for the future results.

Thanks for reinforcing this idea which I'm seeing applies more and more with creating successful strategies.

The data is tricky to analyse, as it's not as simple as trawling through BFBM results (as in UK dog races they only give you a 1/6th of the picture, albeit an important one!).

I picked up on your advice from earlier about how you manage to get those big winners with the long odds. There's a limit to what data is available for the dogs in BFBM compared to horses, so I think we're limited to filtering on the following:

1) Odds ranges of runners we're betting on
2) Odds ranges of runners we're not betting on
3) Spread of odds amongst favourites (e.g. when outsiders win a race are the 1st 3 favs all close and low odds like 2 or 3 ish?)
4) Grade of race (or class and grade in AU?)
5) Distance
6) Day of race (different biases seem to occur at weekends - at least in UK dogs)
7) Time of race (for UK dogs it appears that the afternoons are the best and most reliable times to bet)

Is there anything else which I've missed?

When I've created strategies in the past, I've always homed in on the non-filtered characteristics of the races and invariably I've become unstuck when the variance kicks in. What I like about the sequence is that it's all pretty much random anyway (you could start the same strat at two different times of day and end up with massively different results - instead of a difference of 1 or 2 bets which is what we'd normally get), so if you can get it performing well on one day, it shouldn't take too much work to make it successful every day - at least in theory lol.

[echelon]

Thanks for the hint about checking more deeply the relationship between the runners.

I've realised that I missed two other criteria from my list:

Checking Trap (box) numbers
Checking for possible venue biases

My main point was that we have a finite amount of criteria which are possible to check for dog races in BFBM.

For example, we can't check form, speed, trainer etc.

Anyway, given that we have a limited amount of data which can be analysed to give us that edge, I think you've done an amazing job putting together so many profitable systems from that data!

[danzak]

Need a help with staking and Step Seq for this this data.

I've got only two months of data. I would like to put into SIM mode but need advice from you.

System has been using form filters like Track, Rating, Class, ...

[danzak]

Thank you.
I'll need to think more closely how to improve my SR from 41% to 50% first then use SS.

[Sr8822541]

Hi DN,

First of all thanks for taking your time and sharing with us here.

I have been using BFBM since April and been grinding since then to find out a profitable Strategy. But, I never used forum until two days ago and stumbled upon your posts here and they are a great source of information.

I am a Data Analyst and my complete focus has been on Aus greys. I have gathered over a year data through betfair by downloading daily data files and it took me a lot of time to prepare the data as I have added few more colums which were not in Betfair data.

I usually write complex formulas in excel which has over a year data for different strats to give me an instant big picture of whats happening with a graph and key stats summary, and ran few bots live for a period of time with decent profits but there is always a new pattern or trend which happens during live and it screwed me up few times.

So I have stopped live bots for now and currently been running few more simulations on the Aus greys data. I have tested mainly LAY side so far, esp 1st and 2nd fav with various filtering conditions like box, odds, different recoveries, etc.,

It is interesting you mentioned about RapidMiner which I have used long back, propably I also should try AI and find some new hidden patterns.

With your forum posts, I have now understood Backing side and will probaly test the data for few more patterns as you mentioned so far. Thanks again for sharing.

[Sr8822541]

glad to find someone as interested in this as I am

I do feel there is one thing that differs between success and failure.

When I watched the BetAngel guys video about winning at greyhounds - I knew he had no idea on how to actually win everyday - because he described the plain vanilla things that sound good and would win if you had an unlimited bank and an unlimited time period and ultimately were the same thoughts as everyone else.

It takes an iron focus - it takes discarding being the best form analyst - and it comes down to the simple concept -- using stats alone can you find winners at a strike rate that can make you money ... yes or no.

in terms of back betting that is the simple truth - you will drive yourself batty relying solely on form - you gotta find an edge and exploit it.

Hi DN,

So I have tried a bot with settings and filters I understood so far. I am running from yesterday in simulation, but having difficulties with the bot settings, I am not sure if the basic ones are correct. PFB screenshot, it is following the SQ I mentioned (so no issues here). but not sure why those extra lines are popping up. Also, you have the unsettled bets condition? If yes, whats the value is at? Because of those extra lines, for me it is considering as unsettled markets under Reports and not betting next sequence.

Also for BSP to work correctly for Greys in BFBM, what conditions you have mentioned? I do not want any setting that reveales secret to your strategy like Sequence and selections filters like box, track dist etc.,.

Could you please provide screenshot of settings related to BSP, and other general selection codn, price settings by hiding all other crucial conditions?

Thanks. This would help a lot.

[Sr8822541]

yeah your a bit lost there

Also I do not bet 2 dogs in same race using same strategy

on my other one I built a strategy for each and combined with joint take profit and stop loss

I just found like everyone else betting 2 with recovery - the recovery gets too big -- which is why I focussed so hard on strike rate .. strike rate is everything to be honest.

we do have some people who said they hit 80% with lay strategy - that is great .. I did bit of work on lay betting but nothing compared to back

Thanks for sharing man. I have corrected Price rules now. I had the bet persistence as same.

I just wanted to get the general settings right first. I didn't wanted to place two bets per race, but not sure why it was doing that, again a while later it stopped and just did one bet per race and followed sequence. Not sure why those "BSP order" bets were displaying, and they were counted as unsettled markets, so it screwed up there. Will check with BF team if happens again.

But can you just quickly see the settings I have in pics below. I have masked SQ and other details. but I am not entirely sure if few rules I have highlighted are required for this. If those are extra, just let me know I will remove.

After this I will just move to analysis part and tweak or add filters as I go and will try to achieve the ideal SR. I have over 400k rows of data for Aus GH waiting for me to solve the puzzle.

Thanks again.