My humble trading experience & open questions
- This topic has 4 replies, 5 voices, and was last updated 2 years ago by
.
-
Hello All,
Encouraged by Jebus89 and others who share their trading stories and experience, I too wanted to tell you a bit about my little trading journey “so far”. Initially I started 2016 with Prorealtime to screen shares and funds but discovered quickly that there are incredible possibilities with automatic algorithms. I got hooked with the Pathfinder in 2016/2017 like so many others in the forum back then. I checked the trades in demo for 6 months and it went amazingly well. So I decided to put them live and I made 80% profit within the next half year (thinking to scale up and soon become a millionaire) just to lose all my gains within 1 months as the market experienced a flash crash. Obviously I still had to learn about overfitting and proper money management. (still have to improve to be honest) As I read a lot of threads in this forum and thanks to a lot of recommendations for trading literature did I improve slowly. Big shout out to Jebus89 for recommending Kevin Davey and Jaekle & Tomasini. They published really insightful books with a lot of examples. Especially Jaekle & Tomasini (Or Kevin Davey, not so sure anymore) gave me great ideas on how to properly document the trading strategies. So I established a growing database with all my strategies where I track and document the trades. They also explained the process of IS/OOS and WFT in detail.
As I try to create my strategies with IS and OOS as well as an open untouched time period of 10% to avoid overfitting, I feel still worried that overfitting takes place so I want to see the proper results after I created the strategy. (as real OOS so to speak) I sometimes can’t avoid to look ahead to make sure I am on the right track and the untouched 10% period is as expected. I copy and paste the backtest history plus the real OOS period (after creation) of 6-12 month into my Excel sheets to see whether they perform close to the expectations. (+- 2 sigma of standard deviation) I believe I have roughly 500+ algos I track like that whereas only a fraction performs as expected or slightly better.
The 500+ algos comprise of roughly 30 different strategies where I apply each strategy on different markets (indices/forex/commodities) So some strategies contain up to 20-30 different algos. Are you guys using the same approach? Last time (when I did a big “inventory”) I had circa 80 algos which generated a profit since their creation. But at least 60% of them are not worth the risk as they have either huge drawdowns or an unimpressive Risk/Reward ratio. Others trade very sparsely and I don’t want to put them live until they traded at least 20 times. Or, the position size is too high on some algos and the absolute drawdown and highest loss would be too great for me (and my portfolio) too handle. So most of them stay either in demo or just get tracked over time.
The ones which perform for at least 6 months as expected I will put on demo and/or live. But as it is so often, some of them go directly into a drawdown. Some other directly hit new heights but lose their thrust after couple months and give back all their gains. I also had strategies which performed nicely in demo so I put them live and they started to underperform immediately. Then I put them back on demo to monitor the strategy closely and as soon as I did that it made profit again. So I decided to put them back live, unfortunately with the same result of underperforming. But I verified the trades in live/demo and backtest and they matched. I was just very unlucky with my timing. Those are the situation where I get paranoid and feel a “conspiracy” against me. Do you feel the same sometimes? The question is then when to discontinue the underperforming algos again. For now, I do it when the strategy performs under the 2 sigma standard deviation and when the drawdown is significantly greater than the historic drawdown. But here again, I never know as I hope the strategy will start performing again and new big gains are just around the corner. Unfortunately, we can’t predict the future. When do you stop your algos?
So after I started my own strategies and some from this forum in 2018/2019 I still lost more money and was down 20%. Mostly due to the fact that I started them directly without tracking them as I mentioned before. After I started the tracking process and only put strategies live after at least 6 month of expected returns I recovered my losses and ended up where I started. This is now 2021 after 4 years of learning and trying. Now I am back at square one but hopefully on the right track. I hope from now on I get slowly on the path of positive returns but who will ever know.
I don’t remember where I read it (either in a book or this forum). Someone said it takes roughly 5 years of learning and optimizing your processes/skills to be around net-zero. After that one starts to make small robust gains. I didn’t want to believe it and wanted to have instant gains. (especially after the first success with the Pathfinder) But now here I am 4 years later and that quote is stuck in my head. How true it is/was (at least for me)
Unfortunately, my time became more limited besides work and private life. So the last two years I didn’t create plenty of strategies. That is also the reason why I was rather absent from the forum and I focused more on tracking/ analysing my existing strategies/algos. Using only the algos live which have good stats and trade as expected. Don’t trade the inferior algos just because of hope and wishful thinking.
That is also the reason why I bought the algo package from Prorealalgo to complement my portfolio and make it more robust. I used their demo trial two times before and tracked their posted live trades with my demo trades and they reconciled. So I bought their full package whereas I only trade those ones live which had a reasonable track record. The first two month went really well and the money I spent for the purchase was already recovered. But since then the algos performance stagnated and I am basically at net zero which is quite frustrating. To their defence, they still post the correct picture with their live trades and they don’t try to hide (or sugarcoat) the fact that some of their algos stopped performing as expected. Nonetheless I have to think hard about purchasing their package again. Overall their algos performed alongside the Dax performance with the same drawdown. So technically I could have just bought and hold the Dax at the beginning of the year.
- I included some graphs with my tracking of 2 algos and their standard deviation. (a good and a bad one)
- Also a graph of my performance of 2021 so far. Good Start but since April more or less flat. (including the Prorealalgos)
- A snapshot of my algo tracking/Analysis
Please let me know what you think.
Still open Questions for me:
- For the same strategy, do you create different algos for different markets? How many of them do you trade?
- When do guys stop your algos. Which rules do you have?
- How do you guys track/analyse your algos?
- Did you experience roughly the same things? (Hope, FOMO, Market is against me)
Best regards
Hio, thanks for the shoutout!
I make those “my journey so far” 10000 lines posts here because thats what I would have wanted to read 2-3 years ago! Im very glad to hear that youve found inspiration from one or more of my long posts. It gives me more motivation to keep going!
Ill try my best to answer your questions from my perspective!
1. I think i could have adjusted a few of my algos to different markets, im just very scared of overfitting so i choose the best market and timeframe for my original idea, then i optimize it for that spesific market and once i feel like im done, i run it only on that 1 market. Maybe big mistake from my part, maybe its a smart move. Who knows.. At least for me i know that im running a algo on the best market i could find with my original idea (non optimized code)
2. This was a HUGE question for me in the beginning. Do i stop it when i “clearly see the markets turning down”? No.. i figured out the secret for this question: Make your algos so robust that you never turn it off. If you need to turn it off, you should be able to code in the aprt that turns it off! If its “daily close < MA20” then just code it in. If you think you can outperform your own algo by turning it on and off manually then i dont think the algo is ready for the markets. When i build algos i want to make sure that i dont need to turn them off for the next 3-10 years* (numbers i just made up). I want to make sure that my algo dosnt break my account just because the underlying market is down -20%
3. I analyze them in a couple of ways: I check their performance calender, extract the numbers to excel for each algo and create a combined equity curve to see my worst drawdowns etc. But mainly i analyze each algo individually. If i expect 9/10 algos to have a positive year, then i dont stress it when im down a month or 2 or even 3 in a row. It happens, i still expect each algo to end its eyar on a green candle. If one dosnt and start behaving abnormal ill stop the algo forever unless i figure out what makes it loose and its not overfit.. (loosing money when its Long only and market is up +15% for example..)
4. The market IS against you. You will ALWAYS activate your algos on the “worst timing ever!!!11!”. You will never be able to relax in this game. Its constant emotional rollercoaster. At least its like that for me, and everyone else i talk to! All you can do, is to be as sure as you can become that your algo wont break your account and “press play and stay away”. I think the emotions is for sure what makes people turn on and off their algos and eventually this will make them loose money. If you turn it off because “everything is down”, well after the biggest down days, comes the biggest green days. Let your algo do its thing. At least for me i started with algotrading because i was constantly screwing myself over with fear and greed. Its obviously still a massive part of this game, but not in every trade anymore. that helps me at least.
All of this being said. Keep working on your own algos and yes it takes 1000 ideas to come up with a good one!
1 user thanked author for this post.
Thank you and for sharing your experience and insight.
My turn… may be it’s a sort of therapy writing these lines… I was more of a buy & hold stock person… I started “day trading” with the first lockdown back in March 2020 if I remember well. Obviously looking at static (after the fact) graphs with indicators, I thought it would be simple. Just pick the right couple of indicators and go. I’m still not consistently profitable, or should I say RED on cumulative basis since March 2020. I adapted a few algos from this forum, I made 100% two weeks in a row, and I lost 150% of my gain the 3rd week!
My take from trial and errors. Rule I’m trying to put to myself:
- if you’re not consistently profitable in manual trading, don’t think Algos will help you do better. You need first to understand the markets, price action, support, resistance, some important levels, etc.
- there is no such thing as a “leading indicator”. All the indicators are lagging since they are based on historical data. There are indicators which reacts more or less quickly to price change. So, don’t expect the indicators to save you from mistakes.
- losses are part of the game. It’s like everything in life, or even worse. 75% of CFD individual traders are losing money in Europe. They are not stupid or lacking eduction, or whatsoever… it’s a numbers game,
- never, ever copy a code (indictor, screener, algo) from any internet that you don’t understand in details. You will quickly blow out your account. In addition, some of the codes here are not correct.
- set boundaries: maximum loss per trade, hours of trading, over-night / over-weekend or not (Gaps!), money management (risk % vs account size, which determines with SL, the max position size), invalidation scenario of your trade (cross under previous low or high, close under MA, etc.). If your system doesn’t fit into your boundaries, don’t just increase your SL to increase your win rate. Look at Net Gain / Max Drawdown, if you don’t have minimum 2x, then don’t put the system live.
- spend long hours watching graphs with your preferred indicators when the the market live. You will notice that if for example you put together Signals which mirror the system you want to trade, the Signals will appear and disappear within the same candle. So you system will take a trade based on the first appearance of the signal, but then the market turns and you’re stuck there. So, put some filters and confirmation. UPDATEONCLOSE is an additional security but it will not save you.
- don’t overfit (e.g. CCI>132 or MA[49], etc.). Your conditions were true in the past but market changes continuously.
- don’t overdo separate conditions for Entry/Exit for Long vs Short. If the market was mainly in uptrend during your simulation, then you would have smaller SL for shorts (small pull backs), and vice-versa. However, when the market will turn to the opposite direction, you will be always on the wrong side oof the equation. So keep it consistent.
- if it’s too good to be true, then it’s really too good to be true. May be you have errors in your code such as the same using the same variable twice or anything else.
- chose the right amount of capital, but in any case nothing that can put you bankrupt, and No Credit.
Voilà!
Find exclusive trading pro-tools on 
