Technical investment analysis methods applied to roulette

[birdhands]

No worries about infrequent betting opportunities; we can always increase our unit size if we have an edge.  This ii really great; you guys should all give yourselves a big pat on the back.

[Colbster]

No big pats due yet.  Most recent test had 17 betting opportunities in 116 spins.  12 losses, 5 wins.  Pretty clear signals using the standards I was mentioning earlier, they just did not hit as we would like.  Still plugging away.

[Colbster]

Here is another idea to throw out, as this is the notebook section and it isn't solid enough to put in full systems.  I think it is easier to quantify for a certainty when something is in a trend than to say that it isn't it a trend.  Attached is another spreadsheet (sorry, I know it is a lot, but maybe something will trigger an idea or an improvement that might break this open).  It is another tracker that tells you when not to bet (any time that any of the cells below our spins are red).  When they are all clear, I suggest we bet opposite-of-last.  It does catch some nice streaks of choppiness.

Here are the rules I have put formulas for into the spreadsheet:

1.  I have Excel determine if last spin was high or low.  The next row determines if this was the same or different from the previous spin.  It keeps track of how many in a row of the same we have spun.  Obviously, the higher this number, the more "in a trend we are" and the worse it would be for us to bet OOL (opposite-of-last)

2.  We do not spin if the average streak over the past 6 spins is greater than 2.  This means that we have hit 3 or more twice in the past 6 spins or have had a big streak recently.

3.  Our average is not growing compared to the last spin.  Again, this means that we are not moving into a new trend.

4.  Our standard deviation is not growing.  When we have a shrinking standard deviation, this means that the spins are consolidating.  With the rule that excludes averages over 2, this means that we are consolidating with chops, or no more than 2 in a row.

When all of these indicators show clear, we bet opposite-of-last.  I am trying to bet against random changing it's mind.  I am aiming to bet against a trend when we are not in a trend, or for a trend when we have been in a trend.  I like the though of random continuing it's own trends and not being able to regularly fight against itself.

Ideas are always welcome!

[Colbster]

I am continuing to play around with applying technical analysis methods, despite the mixed bag results.  I agree with those who have suggested that the moving averages methods tend to lag the movements, so I am toying around with oscilators, which are generally considered to be leading indicators.  I have done some testing today with RSI (Relative Strength Indicator) and found it to be lacking for our purposes.  Not only do we have infrequent overbought/oversold indications (>80 or <20), they don't seem particularly good at calling direction.  When we get a hellacious streak of one side of an EC or another, there is no investor sentiment that causes it to come down as a correction.  We are still left largely at the mercy of the market, aka random.

Next up: CCI (Commodity Channel Index), which is one of my personal favorite tools.  Anything that you guys would like me to test, if you let me know, especially if you can point me in the direction of the formula that is used to calculate it, I will be glad to give it my attention.

Thanks for the continued interest and ideas along this discussion line.

[Colbster]

Just tested CCI on a few hundred spins.  It did much better than RSI, giving more trigger readings and performing better.  It does, however, take more tracking spins before the readings are of any value.  I computed the CCI using a 15-period mean deviation instead of the more typical 20-period reading that is commonly used.  Also, for typical price, instead of using (High+Low+Close)/3 for the typical price, I used the SMA for the past 3 spins.  This small test did give me a positive result flat betting, although it did have a losing streak of 5 straight losses.  That said, it also had a winning streak of 9 straight wins.  This seems like it could be of some use, but I think we are getting out on the periphery of what could really be useful to our endeavors.  I would like to crawl back to the mainstream a little bit, which I think is where the moving average crosses are.  Thinking maybe throwing an EMA/SMA cross into the mix to see where we end up.  I will report back.

[Colbster]

I got really promising results using a 10-spin SMA/EMA cross.  As before, I am attaching a spreadsheet for you to test this with your own numbers.  Just finished a 250-spin session that won 11 units flat-betting.  Very solid candidate for a progression, maybe even a martingale of some sort and duration.  The longest string of losses was 5.  Flat-betting, the max drawdown was 7 units and the lowest actual bankroll was +/-0.

Let me know how it works for you.

[Colbster]

Another 193 spins.  +7 at the end flatbetting.  Biggest drawdown was 9 units, lowest session bankroll was -2 units.  There was a single series of 5 consecutive losses.  Of the 136 bets, we would have been +79 with a grand martingale.  I like the short series of losses that I am having so far here.  Will continue to test.

Oh, and for those who have downloaded the spreadsheet, enter the spins across the top row.  The bet is the H or L in the yellow highlighted row directly below your latest spin.

[Colbster]

Short flatbetting session - 39 spins: 27 wins, 12 losses = +15 units.  This was a total of 66 spins, including the tracking spins at the beginning.  I was also tracking the d'Alembert progression to see how it fared against this bet selection.  It resulted in a plus 27, as we would expect.

[marivo]

256 spins, +12, flat. Very stable.

[Colbster]

Just spun 290 and ended with a +10 betting flat.  However, this was a more volatile session than most of the others.  I reached a high of +7 at 58 spins, before experiencing a huge drawdown of 14 to the session low of -7 at 99 spins, which recovered back to around zero very quickly and orderly.  Gradually, the system came back, ending my session at +10 on a streak of 4 consecutive wins.

Still like it, just want everyone interested in this to know that it does have some deviation.  Still, it hasn't lost and has always came back plus some.

[Colbster]

Just had the system breaker series.  I played close to 500 spins.  The high of the session was +7, around 100 spins in.  It dropped down to a low of -18, a draw down of 25.  It then languished between -9 and -18 for an eternity and I gave up.

[Nickmsi]

I tested this system for 7,500 spins of 25 sets of 300 spins each.  Results:


Totals at 1st 100 Spins:   -1    Totals at 200 Spins:   -10   Totals at 300 Spins:   -4


Pretty much break even.  This was all flat betting.

I was betting every time the SMA and EMA crossed and that was a lot of betting.  I was thinking maybe if we reduce the number of bets by narrowing the bet selection, it might help.

Instead of betting every time the SMA and EMA crossed, I bet when they crossed only if their differential was .25.  I tried other differentials, like .01, .50 etc. but most were not any better.
The .25 was area appears the best so far.

Results with .25 differential for 25 sets of 300 each.

Totals at 1st 100 spins:   38     Totals at 200 spins:   112    Totals at 300 spins:  151

It is just a suggestion but perhaps we should look at not only the moving averages crossing each other but that a secondary criteria might enhance the bet selection and improve the results.

Good hunting and keep the ideas flowing . . . Nick

[Colbster]

I concur that adding a second trigger, where both conditions have to be met before we bet, would be an improvement.  I am glad you threw the size of the differential into the mix.  That is not something that I had seen.

Can you look at your data and tell how frequent certain draw downs were?  If we get flat betting results of even, that means we have 1 win for every 1 loss.  That is optimum for the d'Alembert progression, where each chop ultimately earns us +1 unit.  If the draw downs are not typically large (say, less than 10 on average), we could easily weather the downs that would result with a modest bankroll (10 steps in the d'Alembert only costs 55 units).

[Colbster]

I think my greatest interest would be if we could quantify whether we should be betting only when the EMA is on our side (ie H when it is above or below 5) or trending on our side (ie H when EMA is growing or shrinking compared to last spin).

[Nickmsi]

While I was running some of the differentials I did a couple quick EMA and SMA only tests and did not see any positive results, however, I did not use any secondary triggers, ie. differentials just on one side.  Will take another quick look and see if that helps.  Will see if above 5 or whatever makes any difference.

Sorry, I don't keep my testing runs, only the results so cannot tell maximum drawdowns. Will keep an eye out in future runs.