Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now, It is 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)
Webinar: FuturesTrader71 (FT71) on Statistical Analysis for a Broader Edge
Very informative to see broadly how Morad trades. You can definitely see the formal training process coming through in his approach. I like his style of charging a decent fee for the key content but having the fee paid to charity. Nice touch.
Just watched the webinar. I like these ideas not for trade ideas but more understanding market tendencies and then using price to find the trade. Also, as per LBR, studying what happens when the unexpected happens i.e. the market breaking both the overnight high and low, is a worthwhile endeavor. A few questions:
What excel executions did you preform during this webinar? I noticed binning, counting, and a histogram. Anything else?
If you were to do 80/20 breakdown what 20% of excel topics should I learn and master that would give me the best ROI on my time?
If you were to start out at this would you learn excel or R?
I have been a member at Convergent, where I have learned a lot. I'm also happy using the services of EdgeClear, great team that always responds so fast, I'm impressed.
I have a question about sample sizes. In the webinar you state that >250 samples are sufficient.
1. For my journal I am using 30 min data from 2010, about 3000 days.
2. For normal days, I exclude holidays, days after long weekends, 1st and last day of the month. About 2400 days remain.
3. I exclude days with extreme high values (top 5%) for prior day OC & HL and initial balance OC & HL. About 2160 days remain.
4. Then I look which weekday it is, let’s just calculate 2160 / 5 = 432 days remain.
From this point, I apply filters (on pre-defined ranges) on for instance prior day OC/ATR and the day before and on prior day HL/ATR. Depending on the day and the ranges, much less days remain.
Then, when the market opens, I filter on up gap/down gap/no gap. And then, did the gap close during the first 30 minute bar. Less and less days remain. Let’s just assume 45 days.
In the end, I look at the initial balance’s OC value and enter a range to filter on.
And the system presents all historical trades that fit within all these filters, with a win% and the sum of points that would historically have been made.
At this point, normally I end up with 8-25 days remaining, that sometimes give pretty good or excellent stats.
The main question
Is it fair to say that the sample size is 3000? Or would 432 be better?
Or would this be 45 days, as that’s the point before putting the final filter of the initial balance OC range?
I have been looking to find an answer, but never really found it.
The strategy runs since April 2021 is profitable, be it slowly but steady.
@FuturesTrader71, could you please shine your light on this?
I understand you are a busy man, and your input would be highly valued and appreciated.
Thank you,
Deetee
PS. Anybody who has a view on this is welcome to give some input too
Thank you for being a part of both Convergent Trading and @EdgeClear.
The issue with the layering of filters that results in just 8-25 sessions to base hard money decisions on is that the sample is too small. Furthermore, given that you started with 3000 sessions and ended up with data for 8-25 sessions, you basically don't trade that signal 99.2-99.7% of sessions. That is a very rare signal and when it happens, your ACTUALLY probability for a positive outcome is 50/50 still. So if the trade is a loss, you have to sit out a LONG time before another trade shows up. Furthermore, with such a low sample, you have a very low confidence interval to work with.
All in all, if the signal ever shows up going forward, there is so little to lean on to be confident that you would have a positive outcome or a reasonable probability. Does this make sense?
432 is not a bad sample size. Think of samples as tokens that buy you emotional capital or confidence/trust. The more there are, the better. However, there is a point where you just need enough to make the decision to move forward with that setup. You can never be 100% confident.
Since you have run your strategy since April 2021, how many trades have shown up?
Risk Disclaimer: Trading Futures is not suitable for all investors. Past Performance is not indicative of future results.
If you have any questions about the products or services provided, please send me a Private Message or use the futures.io " Ask Me Anything" thread
The following 2 users say Thank You to FuturesTrader71 for this post:
Thank you for your reply @FuturesTrader71. Means a lot to me.
I understand what you write. And it does make sense.
I had 51 FDAX trades in my journal, this were the results till last weekend (IBL = green initial balance, I'm using the first 30m bar as IB):
I understand, that if something only happens 12 times in the last 12 years, it's a long wait indeed.
The thing is, that every day I apply new filters depending on the situation for that day.
Every day is kind of different and gives different statistics.
The historical trades in my database are fixed (long/short, different entry/exit times and 5 different SL), but the results are only counting these historical trades that fit the filters.
Filters like Prior day OC & HL, and the day before OC, gap, gap closed, IB OC.
The output looks like this:
This output example is from today and is not looking very good. Some days have win% >80% with good points per trade. In principal I don't trade if the historical trades are below 8, and with 8/9/10/11 historical trades the win% must be at least 80% (+ good points per trade). But ideally 20 trades or more remain.
I'm following a fixed process on how to use which filters and in which cases, and how to pick the trade from the output (win% is leading).
I'm confused what is the sample size. The 3000 is the total sample, but after cutting it down (till step 4 'weekday' from the example) 432 remain. But then I keep on filtering down, and I'm not sure what I can call sample size.
