Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
Broker: NinjaTrader Brokerage (Continuum Data Feed)
Trading: ES, YM, NQ
Posts: 3 since Jul 2017
Thanks Given: 3
Thanks Received: 7
Hey all,
First post on the forums! I am excited to get more involved here, and so I'll kick this off by introducing my first trading journal on futures.io.
After discretionary day-trading the YM and NQ futures contracts for the past few months with some small funds, I've lost $1.5k. Most of that loss was early on while going through the novice trader phase of learning how to actually follow the rules. Now I have gone breakeven, back and forth, for the past couple months despite focusing on education.
I've recently looked into automated strategies. I used to code before coming to medical school, and I love mathematics and searching for patterns. So I read a few books and will be "incubating" a strategy I wrote coalescing aspects of other people's strategies on this forum and across the web.
I've devised and coded a strategy called "NQRenkoScalper."
Timeframe: intra-day
Type: short-term trend following Contract: NQ futures
Quantity: 2, with different exit parameters for each
Max stop loss per trade: ~$200, although it frequently moves to breakeven
Profit target: none explicitly; exits on weakness - will follow a strong trend to infinity
I did walk-forward analysis to establish best indicator inputs for entry and screening out chop. Then, I tested performance on the most recent, unused data I had, January 2017 - September 2017. I utilized NinjaTrader 8's high order fill resolution using tick data and set slippage assumptions to 2 ticks, as I both enter and exit exclusively with market orders.
Overview
The strategy trades two contracts, and I like the net profit of $3,833.36. Profit factor was 2.66, which I consider too good. I'm testing on data not used for optimization, but still worried about curve-fitting. Max drawdown was -$1,502.36, which for my already-depleted small account is worrisome. Over 9 months, the strategy placed 76 total trades, averaging 0.44 trades per day. My largest winner was $761.36, attributable to the trend-following nature of my strategy. Largest loser was -$183.64. I averaged $175.50 on wins and -$56.32 on losers.
Equity Curve
Nine-month equity curve is shown. The strategy stays completely profitable from the start date of January 1, 2017. However, if trading began around May, peak drawdown would have occurred, creating an initial ~1.5k net loss (red arrow) before the August rebound.
Monte Carlo
Monte Carlo analysis (not shown) shows if trades were randomized in order, the above predicted max drawdown is likely an outlier, with median Monte Carlo max drawdown around -$640 and best-case max drawdown up near -$200. Other results from Monte Carlo show net profit is likely accurate, with peak median net profit after trade randomizations to be around $3,600, with worst-case around $1,800 and best-case up near $7500.
Plan
Starting October 1st, my strategy will run live on a simulated account until February 1st. This is "incubation," which will allow me to sit on the trading idea and set up a more objective evaluation in February before I decide on the strategy's validity. I will examine strategy performance on the 1st of each month and report back here with results. By November 1st, I will also set up comparison strategies with the same code except one that enters randomly, one that exits randomly, and one that both enters and exits randomly. I will use these strategies' performances to further compare to NQRenkoScalper's to ensure I am thoroughly analyzing performance.
That raises a question that I had. If you do get a successful automation strategy to run, how long should you let it run in a simulated mode Before deciding on the results and trading it live? How did you determine your Runtime?
Broker: NinjaTrader Brokerage (Continuum Data Feed)
Trading: ES, YM, NQ
Posts: 3 since Jul 2017
Thanks Given: 3
Thanks Received: 7
I am definitely no expert, but I chose 3 months because
1) I considered it long enough to get over 30+ trades on my system to be able to do statistical tests (which usually require n=30 or higher) and,
2) I am not so concerned that if I do have a "trading edge" with this system, that it will be gone in 3 months. If I chose a year or two years, any edge I have in this system has a higher probability of being invalid after that time.
I'm building strategies for automation as well. It sounds like you're coding them directly, or do you use a platform/software to do this? I'm using SharkIndicators Bloodhound inside NinjaTrader 7 (it only works in 7), which is quite easy in their visual logic builder.
How do you do your backtesting? Bloodhound offers a backtesting option, but I've noticed that applying a strategy to a static chart (historical) has very little to do with running it on a live market. For example, I built a quick strategy that showed a 12+ profit factor using the backtesting option, but when tracking the live market it was only marginally profitable and very inconsistent (big wins and big losses). So now I either run them in sim on the live market, or I do market replay at increased speeds. That has shown to be pretty reliable for testing performance.
I'm most curious to hear (if you don't mind sharing that, I'd understand if you don't) how you've solved for avoiding a choppy market without also missing the good opportunities. It's the part I'm struggling with currently.
Strategy Analyzer (SA) is almost useless with custom bars (and in general to be honest). I use SA is a very limited fashion, maybe to find which moving average value may work best in a strategy set up for example. But I never trust the data fully until I go to the next step below.
Make sure you download historical data for a month or two and run the strategy with Market Replay. I bet your results will be quite worse unfortunately. The strategy may become unstable if you run for long periods of time, so you may want to break the backtest a week at a time, and load your results in a spreadsheet.
After you get a PF of 2.00 or more with Market Replay I would run the strategy against live data for a couple of months. If your PF is still above 2.00 you can consider starting a small account and see how the strategy performs in a live environment.
Broker: NinjaTrader Brokerage (Continuum Data Feed)
Trading: ES, YM, NQ
Posts: 3 since Jul 2017
Thanks Given: 3
Thanks Received: 7
With backtesting, it makes assumptions about order fills that can easily drive up profit factor. If you use limit orders, backtest often assumes that you will get filled when the limit price is touched - not very accurate. Usually you'll want to script in custom logic that order is filled only after one tick past your limit price.
If you use market orders, you need to build in slippage. Without slippage, backtester assumes that you get filled at a favorable price on market orders, which usually you need to jump across the bid/ask to get a fill in live market. And, if your strategy gets you in when price is moving quickly, you might even get 2-3 ticks of slippage, depending on which contract.
To stay out of chop, I use the ADX indicator. But it's relative - unfortunately, it'll still get me in on signals as a trending market is transitioning into chop, and it will keep me out on signals when chop is transitioning to trend.
Yes I've just recently read about this weakness - unfortunate. But, will follow your recommendations here and let y'all know the results.
