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)
Robots Unleashed - a beginners algotrading journey
I read elsewhere and confirmed that TS only offers 6 months of tick data. I wish they offered more. I think what I will do is subject any spot checked tick data strategy with TS to a 6-12 month incubation test to watch it live and record the results.
and after watching the incubation period I will trying a few months trading the micros with it before risking larger capital.
Can you help answer these questions from other members on NexusFi?
This is a wise approach. The problem with tick data, and the reason for the 6 month limit, is the sheer volume of it. If we ever got as much tick data as we need, you would get to a point where you dealing with terabytes of data and eventually petabytes with enough instruments. This is why the big guys have a huge advantage over us. I have years of tick-level data for forex available to me, but it takes an eternity to download.
If you want more tick data, you can always start saving it yourself (moving forward), or you can check out this thread:
I keep getting requests for people to share their GomRecorder data. A few threads have been started on the subject but not many are following thru it would seem.
I am collecting data from IQFeed via QCollector. All data contains bid/ask and is tick …
.
Anyhow, I have the same problem. I am incubating something I developed last year, which is based on tick data, but I need to know if it will hold up long term. My problem on this particular instrument is that there is no micro contract.
Thank you for the referral of this info. I will definitely check it out. I can see how the data may be a problem but If I could have access to the full amount of data for some of the major contracts I could download in 6-12 month segments and backtest manually.
I think I have seen this idea you have worked on in your journal. Wasn't it regarding the VX? How has that done?
My idea is fairly simple. I am looking at a 1440 tick chart of the ES and looking for candles where the open to close is nearly all the range of the candle indicating a strong immediate momentum intraday move. It has done well over the course of the 6 months data I have to test and so far my very limited live data has performed somewhat the same.
I am just recording the end of day data with slippage and commissions added in a google sheet. So far the drawdown is the most attractive part of the strategy.
On another note I have been trying to reoptimize strategies I received as a part of our club bounties and it has been hit or miss (mostly miss ) I just can't seem to find how to get remotely similar results using MultiOpt or manual walkforward. This is disconcerting because my logical mind wants to see some form of agreement from backtesting as my fellow students have gotten. I have attempted to reach out to the developers but haven't been able to connect as of yet.
One great thing about the bounties is you get to see what others have tested and see how they write their code. The ideas are limitless but the time is the constraint.
I was reading today in a forum about someone needing ideas for breakout strategies that pullback for a re-entry and this got my wheels turning.
I remember reading somewhere that a high percentage of breakouts fail. So my idea that I will test tonight is quite simple. I am going to test this idea to try and take advantage of the failed breakouts and I will start with a 1/1 reward/risk ratio because my thinking is if the odds are on my side a 1/1 reward/risk should result in a profitable system.
VX.... wow that was an eternity ago. I had to shelve VX for a while, for two reasons: 1) I was having issues with my limit orders not getting filled and could not adequate troubleshoot or get an answer as to why; 2) Margins were incredibly high for VX at the time probably still are. There is mini-VX now, but I have not entertained that. I will bring that particular system back at some point. Thankfully I have enough good systems running now that I have the luxury of not working on it. But I think about it almost every week. That particular system was built on range bars, so it had the 6 month limitation for tick data.
Your idea is interesting. I believe your idea can be described in candlestick terminology as 'long days' and Marubozu. Here is a little snippet of code I stumbled across this week, that may or may not be useful, where n = number of bars back:
I was puzzling for a long time as to how to measure the current bar against other bars, and this is one solution (and maybe used with a multiple of ATR). It seems so obvious now.
It took me a while to muddle through reoptimization in MultiOpt. It took a lot of trial and error, but I think I was able to closely replicate the results of other's systems. I am still working through the 'back catalog' of the strategies I received from the SF workshop (I think I am done with the bounties I received). I am thinking of writing the re-optimization instructions as a guide, as Dave's video is a bit confusing. I found that you must use the data start date that is shown in the strategy (if it is there). I am guessing that some of these strategies pre-date MultiOpt, so the results may vary from a manual walk-forward or StratOpt. I think that if your reoptimization results are close to the original results, that could be sufficient.
One funny thing happened while looking at one of the bounty strategies: the trades looked very familiar. After reviewing the code, it was almost identical to one I developed a couple weeks prior (mean-reversion).
Let me know if you need help figuring out MultiOpt, as I think I have it working. I learned a lot about optimization and WFA using MultiOpt.
Oh, one last thing: make sure you are using Sunday as your start date for any tick-based chart, otherwise your back-testing will be wonky for that first week. Been there, done that.
You are basically talking mean-reversion here. Larry Connors and Cesar Alvarez have done a ton of work on this and I have been working on this for since the beginning of the year. Here is my one system based on one of Connor's strategies: https://systematicalgotrader.com/2021/02/12/sat2021-02-connors-rsi-1/ I think the exit can be better, but it works for HG and W. I have about half dozen variations of mean reversion, blending indicators and exit variants. Cool stuff. The breakout failure explains why I have almost always been caught on the wrong side of a move in other strategies and trades.
Yes Mean reversion is the theme for sure. I ran my initial idea thru MultiOpt and my preliminary results were promising but the full walk forward left little to pick through. The idea is definitely not scrapped but I thing the entry and exit were a little to simplistic to work long term. I will definitely look at your website to see what I can find out. My idea was short the breakout bar high next bar and buy the breakout low next bar. My first version used stop orders for entry but after thinking it through a limit order would be best thereby eliminating the need for full round slippage. I may not get filled but if price is breaking out already a limit order seems logical.
also for my exit I used a simple stopL and profitT as multiple of stopL. I think I will try an atr stop or chandelier stop with a simple small profit target. Small because I have seen systems with small profit targets work long term. I will report back my findings.
BTW I love multiopt. it has taken me a long time to get used to it but I am now getting in a groove and I love the new features.
I do have a question for you. I was glancing at some of your old journals (specifically the emini dow journal) and I noticed you had said you were having trouble with tradestation. I think you had said you were switching to multicharts or sierra or something like that. What are you using today? did you figure your problems with TS out?
I have problems occasionally with TS crashing and also if I don't go through and refresh my charts daily I get a no order execution issue where the minute I click on my chart the order immediately gets sent. This has led to very late entries. On occasion when I missed the original entry I opt to not click on the chart and just let the order get missed. I currently have a different workspace for each asset class and one for each equity index. I am wondering if having so many workspaces open at a time causes some workspaces to become inactive or something.
My solution is to click on every chart in the morning in order to wake them up so to speak. Maybe this is tradestations way of making sure you are not forgetting about monitoring your strategies. All I know is that this can be quite annoying.
Overall, I am happy with tradestation as I was coming from TOS several years ago which truly sucks in comparison.
I'll never say never but I am probably glued to TS for the rest of my life. Especially now that I have MultiOpt which I don't believe there is anything like it on any other platform.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,049 since Dec 2013
Thanks Given: 4,388
Thanks Received: 10,207
The first few times I won the strategy club (this is pre multiopt) I tried to do the same thing and was disappointed in my ability to reproduce results. Then about two years ago I went back and tried to update my own systems that I entered into and won/passed the SF club. I wasn't trading them and they were 2-3 years old at that point. For at least one of the systems I was completely unable to reproduce the results I got. My own results! I'm not sure why, not sure if I did something wrong the first time, not sure if the data has changed in a way that effected them, but the results weren't even close! And this is my own strategies so I'm not accusing anybody else!
Try trading Bitcoin! With Bitcoin at $60k, a CME future has a notional value of $300k and a marginn requirement of about $125k!
I've tried data mining for patterns like this. Examples being
head < 0.1
or head > 0.5 (ie prices moved substantially higher but then sold off completely)
or body > 0.9 ~ same as your condition50 (ie open to close was very large part of daily range)
or even body < 0.1 (ie open to close was very small part of daily range)
where
head = (high - max(open, close)) / (high - low)
body = (max(open, close) - min(open, close)) / (high - low)
tail = (max(open, close) - low) / (high - low)
so
head + body + tail = 1
I had limited success though. Something I wanted to pursue further but not enough time - as always.
There's a certain Murray Ruggiero Intermarket Strategy that is openly available on the internet that has passed the club multiple times (that I know of) by different entrants.
) I just can't seem to find how to get remotely similar results using MultiOpt or manual walkforward. This is disconcerting because my logical mind wants to see some form of agreement from 
