A staking strategy thats worked for me for months in the wild, I cant replicate in excel

[marketraisen]

Curious if theres a maths reason for what I'm about to describe or if its just me making a mistake.

While trying to help a friend out with some new ideas I thought it would be a good idea to make an example spreadsheet of the model I use for staking in one of my systems.

As I have confidence in gaining a +ve expectation of a return, when I do get a series of loses, I progressively increase my stake slightly in the expectation that a turnaround in fortunes will come (with an obvious protection in place to stop it going out of control with an long series of losing bets, thats happened once so far in multiple 1000s of bets).

for example-
Assume that my current balance is £1000, I stake 1% of that on my next bet and it loses, my new balance is now 990.
Rather than my next stake being 1% of that new balance or annoying level stakes, I stick with the 1% of the recent higher balance (with a lot of VBA f**kery recording in running balances on sheet between races), but then do (1000/990)*10 (the 1% of the 1000) so the next bets stake is £10.20 assuming it loses.
if that loses then my next bet is (1000/979.8)*10, so £10.20.
then (1000/969.6)*10 = £10.31
then (1000/959.29)*10 = £10.42
then (1000/948.87)*10 = £10.54
then (1000/938.33)*10 = £10.66
etc

If I then get a win at 4.0, ignoring commission for now the balance would then move up to £970.31 (for comparison, if I was just doing level £10 stakes the balance would be £970 and at sticking to 1% of current balance it would be £969.72)
my next bet would then be staked at (1000/970.31)*10 = £10.30.

Eventually if the balance gets above the recent high of £1000, my stakes are increasing anyway as I'm still doing 1% of the recent high, using that new total in the (recent high/current)*1% of recent high.

I guess written down that seems quite an aggressive strategy but with the sheet I use it on there has seen great results this past few months, plus it means the sheet can run all on its own with me never having to adjust anything manually. I have paper tested an even more aggressive strategy of always using the recent LOW price in the stake multiplier (ie in the above series every bet with a balance below £1000 would be (1000/938.33)*10 but in the real world it doesnt produce too many different results, just more swings.

..................hold on ill be back in a sec to ask my question

[marketraisen]

Stepping away from the computer for a moment and writing all of that out, I actually realised the answer as to why I couldnt replicate the staking in coin tosses on an example spreadsheet, its because coin tosses are even money shots and typically winners I get on the exchange are in the 2.5-4.0 range.

Actually using the more aggressive strategy of always multiplying the stake using the recent high/recent lows, for coin tosses would work.

Assume you gave +1 for heads and -1 points for tails, every time a running sequence got above 10 and you decided to bet more on the next outcome being tails, increasing stakes the further away from 10 you got, it would actually return a profit every single time. a sequence might look like this.

+10 running total, heads is tossed again....
11 heads
(11/10)*1 = -1.1
12 heads
(12/10)*1 = -2.3
13 heads
(13/10)*1 = -3.6
14 heads
(14/10)*1 = -5
15 heads
(15/10)*1 = -6.5
16 heads
(16/10)*1 = -8.1
15 tails
(16/10)*1 = -6.5
14 tails
(16/10)*1 = -4.9
15 heads
(16/10)*1 = -6.5
16 heads
(16/10)*1 = -8.1
15 tails
(16/10)*1 = -6.5
14 tails
(16/10)*1 = -4.9
13 tails
(16/10)*1 = -3.3
14 heads
(16/10)*1 = -4.9
13 tails
(16/10)*1 = -3.3
12 tails
(16/10)*1 = -1.7
11 tails
(16/10)*1 = -0.1
10 tails
(16/10)*1 =+1.5

Maybe writing all of this hasnt been in vain and I will need to reconsider all of this.

[LinusP]

Why do you do this? There is no edge in staking strategies that replicate any form of martingale.

[Derek27]

As I have confidence in gaining a +ve expectation of a return, when I do get a series of loses, I progressively increase my stake slightly in the expectation that a turnaround in fortunes will come

Hi marketraisen,

I'm well aware that you said previously that you no longer want to engage with me, but I have to tell you that the Martingale system is futile and will never work.

If you're confident that you have an edge and your fortunes will turn, it shouldn't be necessary to progressively increase your stakes. Staking should be intelligently based on the circumstances of the bet you're placing, the number of previous losers are entirely irrelevant.

[firlandsfarm]

Not to answer your spreadsheet question (though I can't see why it can't be performed within the cells … seems to just need a rolling Max column to base your calculation on) and I agree with Derek27 on the need for a loss chasing system however I have often thought that an appropriate staking plan may be able to enhance the profits. But to do so the selections must be profitable in themselves. A staking plan cannot convert selections that make a loss at level stakes into a profit at calculated stakes. But if the selections make a profit at level stakes then if you know your losses will be recovered at level stakes why not increase the stakes when the bank is below expectation and reduce the stakes when the bank is above expectation! Yes, it's the opposite to conventional thinking but if the 'expectation' is accurate it may have legs. The key would be how to calculate the 'expectation' … I don't have a clue, same as I haven't a clue how to set individual odds to make a book.

[LinusP]

Not to answer your spreadsheet question (though I can't see why it can't be performed within the cells … seems to just need a rolling Max column to base your calculation on) and I agree with Derek27 on the need for a loss chasing system however I have often thought that an appropriate staking plan may be able to enhance the profits. But to do so the selections must be profitable in themselves. A staking plan cannot convert selections that make a loss at level stakes into a profit at calculated stakes. But if the selections make a profit at level stakes then if you know your losses will be recovered at level stakes why not increase the stakes when the bank is below expectation and reduce the stakes when the bank is above expectation! Yes, it's the opposite to conventional thinking but if the 'expectation' is accurate it may have legs. The key would be how to calculate the 'expectation' … I don't have a clue, same as I haven't a clue how to set individual odds to make a book.

Sorry wat?

[firlandsfarm]

Sorry wat?

Can you be a little more specific with your question please!

[Derek27]

Not to answer your spreadsheet question (though I can't see why it can't be performed within the cells … seems to just need a rolling Max column to base your calculation on) and I agree with Derek27 on the need for a loss chasing system however I have often thought that an appropriate staking plan may be able to enhance the profits. But to do so the selections must be profitable in themselves. A staking plan cannot convert selections that make a loss at level stakes into a profit at calculated stakes. But if the selections make a profit at level stakes then if you know your losses will be recovered at level stakes why not increase the stakes when the bank is below expectation and reduce the stakes when the bank is above expectation! Yes, it's the opposite to conventional thinking but if the 'expectation' is accurate it may have legs. The key would be how to calculate the 'expectation' … I don't have a clue, same as I haven't a clue how to set individual odds to make a book.

I bank shouldn't have an expected balance ‒ a bank's balance is what it is.

If you increase stakes when the bank reduces it only increases the probability of going bust, and if you reduce stakes when the bank increases you would be wasting your resources.

[eightbo]

Adjusting stake sizes won't change the expectancy of your coin flip.
Assuming a fair coin flip, the more you flip - the closer your heads/tails % will head towards 0.5000000000.

When simulating a coin flip you need to simulate randomness. It's worth noting that:
    - Excel 2007/2003 is terrible at this
    - Excel 2010 saw huge improvements to the RAND() function by switching out the algorithm but it's still not truly random
    - VBA's rnd() uses a different algorithm and is known to use the same seed number each time meaning you'll produce the same numbers

With that said, I made a weighted coin flip simulator using VBA's rnd() which is fun to mess around with (attached).

[marketraisen]

There is no edge in staking strategies

There is but it requires good selections in the first place, I also said "(with an obvious protection in place to stop it going out of control with an long series of losing bets, thats happened once so far in multiple 1000s of bets). " . So I'm not martingaling myself to death. Plus the increases are minimal on their own but make a huge increase in profits over time, if I was chasing 50/1 shots then you could say its crazy but in the theoretical example in my OP, I wouldnt even be doubling my stakes until the balance got as low as 500, from almost 4,000 bets the longest losing sequence has been 27, with the next two losing runs reaching 15.

(though I can't see why it can't be performed within the cells … seems to just need a rolling Max column to base your calculation on)

I already do it=

Code: Select all

Private Sub Worksheet_Calculate()
If Range("BA1").Value <> Range("BB1").Value And Range("H1").Value <> "Suspended" Then
Range("BB1").Value = Range("BA1").Value
End If
End Sub

"ba1" being just =value(c6), the current balance. The suspended part is just to make sure that I'm only recalculating stakes on an per race basis (ie each bet on the same market uses the same multiplier)

Code: Select all

Private Sub Worksheet_Change(ByVal Target As Range)
Dim NR As Long
If Not Intersect(Target, Range("BB1")) Is Nothing Then
NR = Range("BC" & Cells(Rows.Count).Row).End(xlUp).Row + 1
Range("BC" & NR).Value = Range("BB1").Value
If NR > 50 Then Range("BC1").Delete xlShiftUp
End If
End Sub

When the value is updated with a balance change as a race suspends, the value gets transferred to the bottom of column BC, deleting the total on the top of column C. Then I take my small/large totals from the rolling numbers in BC. I experimented with different triggers to check for balance changes when the cells are populated into excel but barely noticed a difference from just using the far simpler suspend on markets closing.

I'm all ears for a less convoluted way of doing this that works...!

I'm well aware that you said previously that you no longer want to engage with me, but

Your input I'm sure as always is appreciated by the forum, we're blessed to have you.

And no, I dont use level stakes but thanks for assuming I do to weigh in with your opinion. Even without the increase/decrease in stakes depending on where my balance is, I use 4 different market indicators to increase/decrease my stake size per bet. Some can be half the size of others, all depends on what the prices have been doing.

[marketraisen]

Adjusting stake sizes won't change the expectancy of your coin flip.
Assuming a fair coin flip, the more you flip - the closer your heads/tails % will head towards 0.5000000000.

When simulating a coin flip you need to simulate randomness. It's worth noting that:
    - Excel 2007/2003 is terrible at this
    - Excel 2010 saw huge improvements to the RAND() function by switching out the algorithm but it's still not truly random
    - VBA's rnd() uses a different algorithm and is known to use the same seed number each time meaning you'll produce the same numbers

With that said, I made a weighted coin flip simulator using VBA's rnd() which is fun to mess around with (attached).

thanks, lets have a look.

[marketraisen]

But if the selections make a profit at level stakes then if you know your losses will be recovered at level stakes why not increase the stakes when the bank is below expectation and reduce the stakes when the bank is above expectation! Yes, it's the opposite to conventional thinking but if the 'expectation' is accurate it may have legs.

Missed this part!
This actually makes sense to me, I guess you could calculate expectation besed on price/strike rate, ie its unusual to have 5 winners in a row, so the expect some draw back?

[johnsheppard]

Seems to me any progressive staking plan will make you more profit faster if you have a positive edge and make you more loss faster if you have a negative edge.

So if you got a positive edge, martingale it up? Right?

[Derek27]

Seems to me any progressive staking plan will make you more profit faster if you have a positive edge and make you more loss faster if you have a negative edge.

So if you got a positive edge, martingale it up? Right?

The Kelly criterium proves that even if you have a positive edge you will lose by over-staking. For example, 6-5 on the toss of a coin and stake 90% of your bank.

[LinusP]

There is no edge in staking strategies

There is but it requires good selections in the first place, I also said "(with an obvious protection in place to stop it going out of control with an long series of losing bets, thats happened once so far in multiple 1000s of bets). " . So I'm not martingaling myself to death. Plus the increases are minimal on their own but make a huge increase in profits over time, if I was chasing 50/1 shots then you could say its crazy but in the theoretical example in my OP, I wouldnt even be doubling my stakes until the balance got as low as 500, from almost 4,000 bets the longest losing sequence has been 27, with the next two losing runs reaching 15.

No there isn’t, you are contradicting yourself even in that sentence. Don’t get me wrong, staking is extremely important as it’s about optimising your profit however anything that involves martingale is complete madness.