Some success with Oscar's Grind on Forex - looking for a coder

I have been trading a system for the past month with some success. I adopted a betting strategy called Oscar's Grind as the form of money management, and I followed the trend for an edge.

It is a mechanical strategy and the gains are small but consistent. This is not martingale but a safer approach with stop losses. I have been trading this manually but it is tedious, as I have to keep setting orders as the price moves.

I hope someone may be interested to code it. No indicators are used, only price. Any interested kindly send me a private message. I am willing to pay a small fee for good coding.

Thanks.

Have a look at this:
http://codebase.mql4.com/source/24568

It's an EA with functions at the beginning that cover Oscar's Grind, "Negative Pyramide", Labouchere, Martingale and "31 System".

Not sure if it works but have fun testing.

Thanks for the link, I've tested those, interesting. I have managed to get my version of Oscar's Grind coded and am carrying out detailed backtesting. Seems profitable but more testing needed to ascertain the maximum drawdown.

I wrote my own implementation in C++ as a science experiment and may port it to MQL if it looks interesting.

My test program just drives it with a coin toss game. My goal is to compare it with Martingale to see how long you can play before you run out of money. I can see the attraction of Martingale since it's the simplest to implement.

Also, the EA at the link I posted could be adapted to run as a script with a coin toss as well. I've heard of the other progression algorithms in the EA but never had a chance to take them out for a drive.

I found the rules to Oscar's Grind here, and being curious, created a XLS (attached) to see whether it was capable of delivering anything other than zero expectancy, as a MM/staking system in itself. Please check that I've implemented the rules correctly.

The XLS shows, that for every possible combination of wins and losses across 1,2,3,4,5,6,7 or 8 trials (i.e. trades), that total expectancy is zero. (I believe that it's possible to prove by induction that the same would hold true for any number of trials from 9 upwards).

You can see this by summing all of the final bankroll values in the gray-background cells (drag the mouse over these cells, and check the Sum=XXXX value in Excel's information panel at the bottom right of the window). In each case, the sum is zero. Of course the XLS assumes that each trade is an independent event, and therefore that each permutation of wins and losses is equally probable.

The result tells me that, unless it can be shown statistically that markets behave in a manner so that wins and losses cluster in a way as to deliver wins for the trader when his position sizes are higher, using Oscar's Grind MM, then there is no benefit in using Oscar's Grind over any other MM system, including consistent sizing (e.g. 1 lot per trade, or 1% equity risk).

I apologize if my comments seem negative, arrogant, or confrontational. They're not intended to be. It's not me that you're up against; it is the laws of mathematical expectancy. During my early years, I spreadsheeted many different types of MM systems, and NONE delivered an edge. I'm trying to save you time in your R&D process, by eliminating approaches that offer no added benefit. However, I fully appreciate that some folk find it helpful to proceed on their own educational journey, in their own way, to experience first hand what works and what doesn't.

FWIW then, here is a summary of the conclusions that I reached:

1. Oscar's Grind is a very benign MM, especially compared to aggressive systems like martingale.

2. If your entries and exits deliver a profitable edge, then there's no need to complicate matters by applying any kind of fancy MM system. You're already on the way to financial freedom. In any case, all MM can do (at least all of those that I've tested) is redistribute wins and losses, with the ultimate result (if tested over millions of trades) as being identical to consistent sizing. Unless you know in advance which trades are more likely to win, and size them higher, then you are merely playing a game of hit-and-miss.

3. Market behavior and probabilities are not going to change because of one retail trader's recent performance (winning or losing streak). Hence there's no justification in varying position size based on recent performance. All losses are equally significant to bottom line, but more recent ones weigh most gloomily on our minds, and an attempt to use unproven MM to accelerate recovery from recent losses represents an emotional response.

4. One needs many thousands of trades to prove whether or not a component of a system provides any kind of an edge. Anything less runs the risk of being 'fooled by randomness'.

Happy New Year, and I wish you all the best with your trading endeavors,
David

In my own testing (with the coin toss game) I compared Oscar with Martingale and straight 1 unit bets out of curiosity.

Bottom line is you can play the coin tossing game 10X as long with Oscar as with Martingale. I wasn't expecting positive expectancy since coin tossing is memory-less and random but the results were interesting. You could play the longest with straight 1 unit betting.

I used this source for my algorithm information.

Steve,

Thanks for the link. I see that it states "However, any time if a bet would win and earn you more than one unit on the series, reduce your bet so that if it won, you would only win one unit." On that basis, there's a small bug in my XLS, as I'm simply staying with the most recent bet size in that scenario, instead of reducing it. I have to go to bed now - will fix it tomorrow. I doubt that it will alter the final zero expectancy result, however.

David

Attached is the revised version of the XLS. There are 3 pages:

1. Oscar's Grind: my original XLS posted earlier. This followed the rules, at least the way I interpreted them, as described in the link in post #5.

2. Oscar's Grind (2): my first attempt at a rule modification, according to the rules in the link that Steve provided (post #6). Then I subsequently realized that I'd misinterpreted the "series" rule.

3. Oscar's Grind (3): I've added a "series" column (aqua background) to keep track of the series P/L, as opposed to the total P/L (which is maintained in the "bankroll" column). The series column displays "no" if we are not in a series; otherwise it displays the series P/L. The "bet size" column follows the rules in Steve's link:
a) if not in a series, bet size is 1 unit.
b) while in a series, if previous trade was a loss, bet size remains unchanged.
c) while in a series, if previous trade was a win, bet size is increased by 1, and then - if necessary - reduced so that (if the next trade proves to be a win) the series P/L can never be > +1. If/when the series P/L reaches +1, the series is over, and the series column displays "no".

In all three XLSs, the total expectancy summed over all possible permutations, over any number of trials (1 to 8), is always zero. This leads me nearer to the conclusion (which I've always suspected) that no matter what MM rules are employed, MM has no bearing on expectancy. In other words, there's no benefit to be obtained from deviating from a flat betting system (e.g. 1 lot per trade, or 1% equity risk per trade). Hence I continue to stand by the conclusions that I listed in post #5.

It's after 2:30 am here in NZ, I'm definitely off to bed this time!

David

Your conclusion regarding expectancy is correct for purely random events such as a coin toss, roulette wheel, etc.

However, it's been established (as long as you don't listen to academics) that markets are not random. Simple candlestick patterns have a hit rate > 50%.
With proper MM these market inefficiencies can be exploited.

hi steve,
happy new year.
what other candlestick-patterns than PINBAR you are using?
I found this pattern the best.
Maybe there is another pattern which is profitable or more profitable?
How are trading the possible patterns?
greetings zackery

I'd rather not do off-topic discussions in this thread.

There's a ton of info elsewhere on this site about candlestick patterns.

Steve,

Lots of technical patterns have a higher hit rate than 50%, but it's proper entries and exits (not position sizing) that exploit these market inefficiencies.

If markets were completely random (like coin toss and roulette), every trading system would eventually deliver zero expectancy, and costs (spread, commission, swap differential, etc) would make that negative expectancy. Therefore every trader would eventually face ruin. But there are many profitable forex traders (myself included, if a two year track record holds any statistical relevance).

My point is that any edge comes from the efficacy of entries and exits, not MM (position sizing). As I said in post #5, unless you know in advance which trades have a higher likelihood of winning, and increase your position size commensurately, no staking system will outperform consistent sizing. The XLS results illustrate that.

Oscar's Grind (and also Martingale, d'Alembert, Labouchere, Guetting, etc) increase stake size not because of the probabilities offered by the market, but because of recent wins/losses by the trader. Hence you would need to prove statistically that wins and losses cluster in such a way that fits the staking pattern of Oscar's Grind. But I suspect that aspiring traders attempt to use unproven 'recovery MM' in the hope that it will somehow accelerate recovery from recent losses. But if it equally accelerates the path toward ruin, then its use is mathematically unjustified.

Players eventually lose at games like roulette not because their staking system (MM) is inferior to the house's, but because the zeros on the wheel make roulette a negative expectancy game. Trading is the same; as I said, it's the entries and exits that determine the expectancy.

Blackjack is mathematically beatable because card counters can increase their stake when there are more high cards remaining in the deck, thereby turning a -EV game into a +EV one. They know in advance which of their 'trades' have a higher likelihood of winning, and increase stake size commensurately.

Hopefully I've explained myself better this time. I also posted on a similar topic here.

David

It's an old argument neither of us can resolve here. Van Tharp insisted you could have a positive expectancy system with random entry (in a market, not a coin toss) and proper MM.

I admit I have yet to prove this nor have I seen a reproducable setup that proved it.

van tharp can suck it.

what is considered a 'random entry'? is it a randomly timed entry? a randomly timed and random direction entry? a random direction but specific timed entry?

it's nice that talking-heads can sit on the shoulders of some hedge fund guys and walk away and write a book about all they 'know'. it's laughable that these psyche guys get to see supposedly profitable traders and then walk away telling everyone how easy it all is, and for only 20 sessions at $x a session plus buying a copy of my latest book. then the talking-head can 'play trader' and pretend he knows what he is doing.

bottom-line is that oscar's grind is just compounding on steroids. you aren't waiting for a few hundred trades of profit to then bump your size, you are just doing it straight after each win. it's just trying to really press into potential win streaks, while trying to avoid enhancing losing streaks.

it's a bit of an anti-martingale, and is just a spin on averaging-up. it's just about the most aggressive averaging-up method you will come across apart from going all-in after a win.

on a 50/50 game all you get is swings that push you towards profit over time, but also still suffers potentially large drawdowns as your bet size increases after wins. much like martingaling it is best attempted with a very small starting bet size and a nice chunk of disposable play money to back it.

Thanks for all the thoughts on Oscar's Grind, and to Hanover for the detailed spreadsheet.

Oscar's Grind can be compared to anti-martingale, but is less aggressive. I take many small losses, while waiting for the 100 to 200 pip run that happens quite often actually. When that tend comes I cover all my losses and make a little money.

My entries are not random and are very simple. Simply buy or sell in the direction of price movement at certain levels. So far my EA has shown a 10% drawdown over a one year period, with 20% net profit. However, I'm still trying to backtest with 99% modelling quality over 3 years.

Will see how that goes...

Just to be clear about where I'm coming from.

1) I don't believe that random entries are workable.
2) I really don't like Martingale very much but find these mathematical ideas stimulating.
3) The experiment I was doing involved coin toss simulations. I was not looking for a "system" that could squeeze magic profit from something with no statistical edge.
4) The coin toss experiment was purely for the purpose of *comparing* how Oscar's Grind compared with Martingale and straight 1 lot betting.

Based on coin toss simulations, Oscar's Grind allows you to play 10 times longer than Martingale. So if you're looking to stay in the game longer Oscar's Grind is "better" than Martingale.

The other measurement I made was the maximum balance attained during each run (until you bankrupt). Oscar's Grind performs much better on average than Martingale or straight 1 lot bets in this regard. So if you're looking for a shot at doubling or tripling your pot, Oscar's Grind is the best of the 3.

When the egg nog wears off I'll share some of the numerical statistical results I found.

What do you mean by "MM"? Position sizing; or both position sizing and trade management (exit)?

I follow Ryan Jones' approach: MM incorporates position sizing and method (fixed frac, fixed ratio, optimal-f, etc), but not trade management.

Perhaps you define "MM" differently, and our differing interpretation is causing confusion in our discussions.

I've read Dr Tharp's Trade Your Way to Financial Freedom. He talks about Tom Basso experimenting with coin toss entries, and dicusses the results. I posted on this topic here.

Anyway, I've posted the conclusions that I've reached, namely that position sizing has no bearing on expectancy. Casinos count on that fact, i.e. that negative expectancy games can not be beaten using any kind of staking system. It's time for me to move on.