I'm going to cover work I've done in the last 4-6 weeks on Renko-based systems. I'll be sharing my three best systems from my work. In this post I'll cover the first and most simple of the three, 379. I'll start off with results
379 TF 12-tick Renko (8/11-3/12)
Profit factor 2.40 / 2.48 Long / 2.33 Short
Net Profit $189,650.00 (1,406.95%)
MaxDD $-1510.00
Max time to recover 6.43 days
Average trade $71.92
Ratio Average Win/Loss 2.72
# of Trades per Day 11.62
Average time in Market 39.2 minutes
379 TF 9-tick Renko (8/11-3/12)
Profit factor 2.20 / 2.24 Long / 2.16 short
Net profit $187,930
MaxDD $-1,240.00
Max time to recover 4.07 days (!!!)
Average trade $49.43
Ratio A Win/Loss 2.52
# of Trades per Day 16.75
Average time in market 23.6 minutes
379 just uses VROC to dictate when to enter trends. Having specific inputs/coefficents in the VROC with different brick sizes really improves data almost across the board.
The aim in building a Renko system is to beat the statistics of a regular Renko system (buy on green, short on red). In summary, here is what I've noticed with Renko systems in general:
As brick sizes increase, profit factor and average trade increase, and max drawdown and net profit decrease. However, the maximum time to recover is a parabola, with a minimum at a brick size of 9 ticks for most contracts (under 5 days).
By using a "filter", one can reduce the number of bad trades (in terms of % profitable, average losing trade or both) - this allows a smaller brick size to be used, meaning that trade frequency, profit factor and consistency dramatically increase, while max drawdown and max time to recover decrease. In effect, it brings the pro's of both large brick and small brick Renko systems together.
It's much more difficult to "filter out" bad trades than filtering out really big losing trades. Life would be much easier if we could filter out bad trades over big losing trades, or even better - both.
Volume-oriented algorithms seem to be the most effective. These work by identifying trend and range periods, since Renko charts like to work really well in trend periods (obviously).
I attached the two systems in Ninjascript and C# (you will need some other indicators though).
I have not gone live with any of these systems (real cash). I've had success with market replay data and live data on a sim account with my broker (a real trading account on their servers, without any cash, for maximum realism). r379 is the standard limit order version, b379 is the market order version. If you backtest in Ninja, you'll need to use b379 with one tick of slippage, and you'll need to be certain the backtest does not play games (other Renko charts that have been widely discussed here help fix this).
I really hope we can get some great ideas swinging around here - in the end we can all take home a profitable system (plus some fresh ideas) and successfully draft these big trucks on the exchange. I'll be working on sharing my other two pieces, and some other code too to hopefully get some more of your mental inspiration going.
Thanks for letting me participate on this board you can work on the Street, trade for your own startup, get a degree from Harvard or Stanford but there still isn't anything quite like retail trading, for better or for worse. Having a paid board is a great idea, keeps the loons out and prevents others from leaking code (most of the time). It's great to know we got each other's backs worldwide ;).
Edit: Here are the settings for the system. You can use the default settings (don't work well on anything but the lousy NT7 Renko) or opto your own.
VoN is the VROC period
Co is the VROC smoothing constant
V1 is the minimum VROC level
Q is the quantity (contracts)
MoN should always be left at 2