Let’s talk about Backtesting

[Zigabel]

Back testing is a technique traders do employ to test their new strategy or in some cases, retest their former strategy to tweak it and see how best it works without actually going live with their trade at the instance, This technique although has helped some traders but at no points guarantees that your strategy becomes a Holy grail because it possibly gave some really good results throughout the test, this is because some live events may not be covered in that backtest, such as the effect of news on trades and other phycological and emotional factors.

Back testing does help but doesn't and never guarantees success, always make sure to keep your emotions and risk management right, it helps even more because Back testing is more of a technical test than a psychological one.

[AVE5]

As a skilled trader in order to survive the market you have to come up with a strategy that suit you the most , and backtesting is what will help you figure that out , without wasting any valuable resources.

Is it really necessary that trader's must work with strategies that suits them? Or how's this suitable strategy identified? I want to know if maybe I'm missing something.
Well, I think the only strategy that's required which can be assumed to be suitable is the risk management in reference to risking what we can afford to loose in regards that trading can be as risky as our predictions can't be guaranteed.
Other than that we should be more concerntrated to dive into strategies that we think that can bring us success.
I'm trying to say that instead of longing for strategies that suits us, we should be more concerned to discover strategies that suits the market so that we can keep long term success.

[M47AK16]

Just as most users have said, back testing don't guarantee a future profits. It's the same way in every field. You practice simulated reality to have a clue of  how effective a product or idea will look like with already known phenomenon in reality. That is what pilots do before they start flying, even soldiers play war games before engaging in war, but until a product is battle tested nothing can be assured. That is because it cannot produce every conditions all the real world.

You just have to be as dynamic and free flowing as possible to survive in the market, because nothing is certain. Historical study will not save you when you encounter sudden volatility and spreads that have never occurred in the past in live market condition. No one is saying back testing is not important but it has its limitations.

Back testing from the word back, is about older data. Though some one can rely on pattern, hoping history repeats. Maybe if we combine future data, that makes our strategy stronger. Apart from what you said, using simulation can also familiarize us on the features. So it still gives some confidence.

When we are in practice mode, we are calm but once we are now in the real thing, we may feel pressured and that is another reason that affects our performance. As long as we use old coins, then all possibilities must be given out already. An old coin like Bitcoin won't likely crash too or experience a technical issue, which are among of those unexpected events.

[Ndabagi01]

Back testing from the word back, is about older data. Though some one can rely on pattern, hoping history repeats. Maybe if we combine future data, that makes our strategy stronger. Apart from what you said, using simulation can also familiarize us on the features. So it still gives some confidence.

When we are in practice mode, we are calm but once we are now in the real thing, we may feel pressured and that is another reason that affects our performance. As long as we use old coins, then all possibilities must be given out already. An old coin like Bitcoin won't likely crash too or experience a technical issue, which are among of those unexpected events.

Taking trades in a demo account and when already funded are two different scenarios and you can never compare them with each other. In trading, no matter how risk conscious you are, you can never compare yourself to have a clear mindset when trading a real account as the way you do when trading an account that is not real.

Performance in trading has a lot of etiquettes that when not followed, your performance level will be low even when you have the real money to trade. Backtesting will help you to test the strategy to choose the best ones but that does not make you to be free from the emotional feelings that come with trading that will also affect your results.

[TedMosby]

I was actively testing trading strategies with OpenClaw + DeepSeek V3.2 before. I am not a pro trader, I don't even know how to read chart patterns. So, I just asked my AI agent to find and try as many strategies as it can that can give me a decent profit. I used BTC historical datasets from 2018 to 2026 on Binance that I found on Kaggle [1] to backtest the strategies.

It's all based on technical analysis, no fundamental analysis, no macro related analysis. The backtesting I did was for futures trading using 15m, 1h, 4h, and 1d charts. Pure algorithmic trading, not LLM based trading with reasoning. The AI was only used to find and test the strategies.

I don't have the documentation anymore. But I still remember the result in general. It surprised me a bit. For that long period, 2018 to 2026, most of the strategies were only able to make 1 to 5% a year.

There's one strategy that was able to make 400%+ profit. But still, it took a long period of time to achieve it. It's not what I was looking for. Maybe my goal is too high because my reference was some videos that I watched on social media showing people building an AI trading agent that can generate a very huge profit in a short time, which now makes me think they are fake videos to get many views and engagement. It's something that is easy to make.

This, for example, only took 5 minutes for me to create this profitable, sophisticated, and FAKE trading terminal. With extra made up narrative, some people might trust me that I successfully created a profitable AI trading agent.

https://claude.ai/public/artifacts/9e319945-4dd5-4963-9c6e-c9a30258c693

Preview:


Maybe there are some people who successfully built one, the real one. The failure I experienced could be caused by me too, skill issue.

Source:
[1] https://www.kaggle.com/datasets/novandraanugrah/bitcoin-historical-datasets-2018-2024

[YOSHIE]

Many traders overlook the importance of backtesting.

I have several friends who can be said to be seniors in trading in the Crypto market, I want to say they once told me the Backtesting trading strategy, but I monitor how they carry out crypto trading specifically, they still use stop-loss and take profit methods in trading activities, but for so long I have monitored the way they carry out trading, I have not seen them effectively in making a profit.

Even though they say the Backtesting method can be optimal in helping crypto trading and can also reduce emotional risks, I have not really seen that the Backtesting strategy is a good method and works perfectly in crypto trading.

[#CapsLOck]

While backtesting is great, there is one issue it does not account for: overfitting. If a trader continually tweaks the parameters until the past performance is etched in his mind, then maybe he is just playing into the past and not discovering a strategy that will work under other market conditions. I believe it is beneficial to preserve some of the historical data as is and apply it only after the strategy is set. That provides a feeling of how far the strategy can truly go, or whether it's just learned a tactic from the previous game.

[Peanutswar]

Even though the strategy right now im using is working in my side still I do keep having a backtesting or basically testing the set up if there are some flawless, also if I discovered a new strategy came from other traders and wanted to check if this will be applicable to my current set up is I do often a testing though considered this as a part of learning at the same time. Most of my testing is the losing trades because I need to check what are the possible signs and pattern reason for reversal of market.

[EarnOnVictor]

While backtesting is great, there is one issue it does not account for: overfitting. If a trader continually tweaks the parameters until the past performance is etched in his mind, then maybe he is just playing into the past and not discovering a strategy that will work under other market conditions.

That is it. Backtesting is about testing what has happened in the past in relation to how the strategy behaved. But the big question is: is the history being studied by the strategy the true reflection of what happened in the past? The answer mostly come out with a No because most strategies rely on the flipping of the market prices, as the market price toss them up and down, and what should have reacted abnormally immediately or repainted would appear as a perfect latch in backtesting. This can't reflect the true actions and decisions of the strategy; the hence the reason why backtesting has never been a good way to test a strategy.

[savetheFORUM]

Yes, backtesting is really important but I think most of us, who have been into trading for quite some time now are already aware about backtesting. We will never make our strategy go live without actually testing it. I have spent months in testing my strategies before actually investing any money. In trading, we have to be utmost careful and testing our strategies before hand just gets us prepared to see what is coming next. Without testing strategies, we would simply be risking all.

You don't need any website for backtesting as long as you know what you are getting into. Yes, your referred website might make backtesting a bit easier and I think newbies should definitely take a look into it.

[Hamza2424]

To be honest bro, testing on live accounts seems like a big deal but it is not because backtesting is not just for newbies, it can be used even by experts to test their strategies and different data and formulas to test if they would have seen the market from this side maybe they could have made a profit, but I also have another theory that it only works in the shortest time frames like for day traders using leverage.

And live trading is not a big deal because newbies can start with a small amount as low as $5 to $10 and they can learn everything. Even before testing it out on charts, I was one of the people who tested everything on live charts but still later had to use backtesting for several things. It is like a playground where we make strategies and see if we survive or not.

[GbitG]

I think OP means that you should journal like record the trade mean "Journal"..... Personally I write down every single trade even all of the rules for my trade and to be honest this is the only way to improve his trading capacity because through this you will be able to understand that how actually trading works and how I am doing it.....

Yes, I know that is what he posted about, but what I meant is just that I do not agree with him so far I have it is automatically written on the trading history of the exchange, but which I will not have to also go through because I can easily know if my strategy is working and maybe to further with the strategy or not.

I got it Dude....!!! 

Trade /Position history is different wether you keep it or not like it's up to you but here we want to discuss that "Trading Journal". Like all the traders around the world the just mark up the entire POV of every single trade and that's what I want to tell you. Because every trader should have the desire to become better than the previous day. So that's why Journal help you to improve his mistakes wether it's about strategy or it's about discipline or even physiological effects too.....

On Exchange you have only tarde history (entry, exit, profit/lose). Despite you will never see the logic behind tarde, your discipline, emotions, risk management instructions or even your mistakes too. If you have good memory like if have remember all of your tarde element then it's okay but in my opinion Journal will good for your improvement Bro......

[henmark]

While backtesting is great, there is one issue it does not account for: overfitting. If a trader continually tweaks the parameters until the past performance is etched in his mind, then maybe he is just playing into the past and not discovering a strategy that will work under other market conditions.

That is it. Backtesting is about testing what has happened in the past in relation to how the strategy behaved. But the big question is: is the history being studied by the strategy the true reflection of what happened in the past? The answer mostly come out with a No because most strategies rely on the flipping of the market prices, as the market price toss them up and down, and what should have reacted abnormally immediately or repainted would appear as a perfect latch in backtesting. This can't reflect the true actions and decisions of the strategy; the hence the reason why backtesting has never been a good way to test a strategy.

Majority of times, history does not repeat itself when it comes to trading with cryptocurrencies. Even if our strategy works fine on the historical data, it does not mean it would work on the current charts. This is why I think back testing is not enough to test our strategy and enter the markets. We also need to analyse the current trends and only than try testing our strategies. I would prefer paper-trading in order to test my strategy on the current trend. I get way better confirmation there instead of tracking historical data.

For some, back testing might still be a useful guide in order to get started so we can't really bad-mouth back testing. There are a number of folks newly entering crypto markets and back testing can be useful for such users.

[I_Anime]

I was actively testing trading strategies with OpenClaw + DeepSeek V3.2 before. I am not a pro trader, I don't even know how to read chart patterns. So, I just asked my AI agent to find and try as many strategies as it can that can give me a decent profit. I used BTC historical datasets from 2018 to 2026 on Binance that I found on Kaggle [1] to backtest the strategies.

It's all based on technical analysis, no fundamental analysis, no macro related analysis. The backtesting I did was for futures trading using 15m, 1h, 4h, and 1d charts. Pure algorithmic trading, not LLM based trading with reasoning. The AI was only used to find and test the strategies.

I don't have the documentation anymore. But I still remember the result in general. It surprised me a bit. For that long period, 2018 to 2026, most of the strategies were only able to make 1 to 5% a year.

There's one strategy that was able to make 400%+ profit. But still, it took a long period of time to achieve it. It's not what I was looking for. Maybe my goal is too high because my reference was some videos that I watched on social media showing people building an AI trading agent that can generate a very huge profit in a short time, which now makes me think they are fake videos to get many views and engagement. It's something that is easy to make.

This, for example, only took 5 minutes for me to create this profitable, sophisticated, and FAKE trading terminal. With extra made up narrative, some people might trust me that I successfully created a profitable AI trading agent.

https://claude.ai/public/artifacts/9e319945-4dd5-4963-9c6e-c9a30258c693

Preview:


Maybe there are some people who successfully built one, the real one. The failure I experienced could be caused by me too, skill issue.

Source:
[1] https://www.kaggle.com/datasets/novandraanugrah/bitcoin-historical-datasets-2018-2024

I have also tried trading with ai scanner. Though I wasn’t the one that created it , I had to pay just to make use of it then . My experience wasn’t a good one at all , end up losing more than the money I use in registration ( In order to make of the bot ). Many was actually testifying on how good the bot is , but I personally couldn’t keep up with the bot . I even tried signals group which was another big mistake I made then , so I had no choice than to learn the skill first myself. Since then I’ve seen better results compare to the signal group and the bot . But if you are able to create a good trading bot , it will be nice too especially in given market bias .

[nelson4lov]

~Snipped

There's one strategy that was able to make 400%+ profit. But still, it took a long period of time to achieve it. It's not what I was looking for. Maybe my goal is too high because my reference was some videos that I watched on social media showing people building an AI trading agent that can generate a very huge profit in a short time, which now makes me think they are fake videos to get many views and engagement. It's something that is easy to make.

The reality is that anyone who says something like that only wants to sell their products or services to unsuspecting, desperate people. The current AI models are not good enough to have an edge on their own because if that were true, there would be so many millionaires already from AI trading agents, but there are no such millionaires yet (or none that we know of). Anyone who manages to do it would keep it a secret; otherwise, they would just dilute their edge. As a matter of fact, I don't think the top AI providers will even let their AI models be profitable.

[BitMaxz]

I have also tried trading with ai scanner. Though I wasn’t the one that created it , I had to pay just to make use of it then . My experience wasn’t a good one at all , end up losing more than the money I use in registration ( In order to make of the bot ). Many was actually testifying on how good the bot is , but I personally couldn’t keep up with the bot . I even tried signals group which was another big mistake I made then , so I had no choice than to learn the skill first myself. Since then I’ve seen better results compare to the signal group and the bot . But if you are able to create a good trading bot , it will be nice too especially in given market bias .

I don't believe to AI too much when it comes to trading without trying them out manually on my own.
AI won't guarantee you a positive result; the market is always changing  if you see the high profit on those AI backtest using them in live trading would give you different result.
That is why it is good if you manually backtest a strategy like the replay from TradingView or free from GoCharting. This manual backtesting will give you a shortcut to test a strategy that you can use in live trading. The result isn't instant, similar to AI, but you can get the fast result after you set your trades and fast forward to get the result fast.
At least you also see some patterns that you need to remember. If you did make some mistakes, you can get results fast and learn from those things that you can apply to your live trading.

This one will increase your chances to get a positive result more than using AI because you can avoid those mistakes you made during backtesting because most of the strategies out there work if you know how to optimize them.

[Zigabel]

I have also tried trading with ai scanner. Though I wasn’t the one that created it , I had to pay just to make use of it then . My experience wasn’t a good one at all , end up losing more than the money I use in registration ( In order to make of the bot ). Many was actually testifying on how good the bot is , but I personally couldn’t keep up with the bot . I even tried signals group which was another big mistake I made then , so I had no choice than to learn the skill first myself. Since then I’ve seen better results compare to the signal group and the bot . But if you are able to create a good trading bot , it will be nice too especially in given market bias .

One thing we need to understand with the AI revolution is that it is not perfectly an independent system, you will need to get it assisted to get the best of it, it can only run on the information available to it, and that makes it has its own lapses that can be a problem at some point, it is also very possible that its developers could be using it as a way to take advantage of its users to earn because they do understand that there is actually no guarantee, but because people wants to cut corners, they could sell those ideas to them and make some money off it, because tell me why they would want to expose their very profitable AI model out there for some coins when they can make tons of dollars form it.

It is always best to learn the skill for yourself and employ that which you have learnt in a way that will be tailored to suit you than the pattern where you will be so dependent on the knowledge of others which isn't reliable at all times. Signals can be good and also has got its down side but with your knowledge, you will be able to utilize opportunities better even as signals or that which you choose to trade on, and you take responsibility better with your knowledge than you would with signals.