YT YouTube2Text

YouTube2Text → Transcripts → I Backtested the Original Turtle Soup Strategy for 15 Years (ICT's Mentor)

I Backtested the Original Turtle Soup Strategy for 15 Years (ICT's Mentor) — Transcript

by Strange ICT · 2,495 words · 390 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Hello guys and welcome back to the
0:01channel. Today we're going to break down
0:02the original Turtle Soup strategy from
0:05the famous book Street Smarts written by
0:08the legendary traders Linda Raschke and
0:10Larry Connors. This book has influenced
0:13millions of traders and many of its
0:15ideas were later repackaged in different
0:17forms. It is basically a trading manual.
0:20It doesn't bury you in psychology or
0:22unnecessary jargon. It gives you clear
0:25setups, rules and examples. And we're
0:27going to reconstruct Turtle Soup as
0:29faithfully as possible and backtest it
0:32across 15 years of historical US 100
0:35data. So first we will test the original
0:37strategy, then we will test different
0:39parameters, values and different ways of
0:41managing the trades and finally we will
0:44see whether we can improve it using AI
0:47without over fitting it. Basically
0:49without changing the rules so many times
0:52that they only work on the past. All
0:54right? Now to do that I'm going to use a
0:56strategy architect, an AI powered
0:58research system that I built for this
1:00exact process. I spent a lot of time
1:03building this engine and you're going to
1:05see how it works. Before we start guys,
1:06make sure to check the current discount
1:08bonus is running. They offer three
1:10options, the classic evaluation, a
1:12one-time fee evaluation with no
1:13activation fee after passing and a
1:15straight to funded option. You can use
1:18the coupon code STRANGER for the best
1:20available discount. And now let's get
1:22into it. So the main question we're
1:24answering today is not did Turtle Soup
1:26ever work because that would be an
1:28insult to a legendary trader like Linda.
1:31The real question is does Turtle Soup
1:33still work today? This strategy guys was
1:36published almost 30 years ago and since
1:38then millions of traders have read the
1:40book, applied its ideas and repackaged
1:43them. Markets also change when more
1:45people discover the same opportunity,
1:47that opportunity can become smaller or
1:49even disappear over time. So we're going
1:51to find out whether the original idea
1:53still contains an edge in today's
1:55market. I'll use this report throughout
1:57the video. This version is only 16 pages
2:00because I created it mainly as a visual
2:02guide for the recording. If you want a
2:04complete technical report with exact
2:06methodology, tests, and results, all you
2:09have to do is to subscribe and leave a
2:11comment down below and I'll send you the
2:12extended version so you can study the
2:14process and apply it to other strategies
2:17so you can find
2:18um profitable edges using it.
2:21All right. Now, let's look at how the
2:23Turtle Soup worked. For a long trade, we
2:25begin by finding the lowest price from
2:28the previous 20 trading days.
2:31We then wait for the market to break
2:32below that low. Okay? Normally, a new
2:3520-day low looks like a bearish
2:37breakout. Okay? Breakout traders might
2:40may start selling because they expect
2:42price to continue lower. Turtle Soup is
2:44betting on the opposite. If it waits for
2:47that breakout to fail and if price
2:49breaks below the old low and then
2:51reverses back above it, we enter long
2:54using a buy stop.
2:56Now, the previous low must be at least
2:58four trading session old, okay? And once
3:01it breaks below it, we place again a a
3:03buy stop five to 10 ticks above the
3:06level.
3:07If price keeps falling, we don't enter.
3:09The order only activates if price
3:11reverses and comes back inside the old
3:13range.
3:14And that order is valid for the current
3:16session only. We don't leave it active
3:17for the next day or try to re-enter on
3:20next day.
3:21Our initial stop loss goes one tick
3:23below the current session low. If the
3:25trade moves in our favor, we trail the
3:26stop loss to protect the profit. Okay?
3:31And there is only optional re-entry
3:33rule. If we got stopped out on day one
3:35or day two, we can enter again at the
3:37original price. For a short trade, we
3:39simply reverse everything.
3:42So, the entire strategy is built around
3:44a failed breakout. You can see it
3:46clearly on this example from the book
3:48from December 1995.
3:51We have a nice range right here. So, as
3:53you can see, on one
3:55price takes a high. This high, if we
3:58scanned the last 20 days, this is the
4:00highest high.
4:01Price takes it. Once we traded above,
4:05we place a sell stop at that level, and
4:08price will take us with it. And as you
4:09can see, this was profitable,
4:11and you start trailing the stop.
4:14Same thing here. Price takes the low of
4:16the last 20 days.
4:17Takes it. We place a buy um stop. So, if
4:21price reverses and came back to that
4:22level, it's going to take us with it.
4:24So, we have two winning trades here.
4:27This one was a loser, as you can see. We
4:29takes the high here with this five
4:31candle, but it got stopped out. All
4:34right?
4:35This one was only uh was also a winner.
4:38As you can see, we took the this low. If
4:41we scan the last 20 days, this is the
4:44uh lowest low, and it happened
4:47um
4:49at least four sessions earlier, as the
4:52book suggested. This was only This was
4:54also a winner.
4:56Uh my initial observation is that this
4:58idea may work better in ranging or less
5:00directional markets. If NQ is trending
5:03aggressively higher, for example, the
5:06long setups may work while the short
5:08setups repeatedly fight the overall
5:10trend. But, this is only an observation.
5:12We're not going to build a strategy
5:13around it unless the data confirms it.
5:17For this research, we're using US 100
5:19data from 2012 all the way to 2026,
5:23divided into three periods, as always.
5:25I'm not going to go through the splits.
5:27It's very obvious. You can see it here.
5:29The first split is where we are allowed
5:31to develop and find variations and
5:34hypotheses and so on. And then we're
5:36going to freeze them and test them on
5:38the other splits. That's it. But, this
5:39separation is extremely important. But,
5:43uh we need to make one important
5:44adjustment, which is normalizing the
5:46trading cost. Extremely important
5:49concept that can easily kill profitable
5:52strategies while back testing.
5:55So, what is normalizing trading cost?
5:57It's easy. All you have to do
6:00or let's say that the account we're
6:01modeling, okay? An FTMO account for
6:04example has 1.45 point spread on US 100.
6:09Okay? Today US 100 is trading around
6:1129k. So, 1.45 point is very small cost
6:16relative to the price of the index. I
6:19mean your stop loss on NQ is is going to
6:22be like 30, 40, or 50 points. So, the
6:25spread is not going to bother you at
6:27all. But in the earlier part of our
6:30data, 2012 all the way to 2017 or
6:33whatever,
6:35the index was much cheaper. If we charge
6:37the same 1.45 point when the index is
6:40trading at 10k for example, that cost
6:43becomes almost three times larger
6:46relative to the value of the market.
6:49Meaning the stop loss is not going to be
6:5150 or 60 points like today. It's going
6:53to be like
6:5520 or 15. So, the spread is going to eat
6:57a lot. So, that would unfairly punish
7:00the older trades.
7:02Okay? So, all you have to do is we scale
7:04the cost with price of index. We're not
7:06removing the cost or trying to make the
7:08strategy look better. We're simply
7:10keeping the relative burden consistent.
7:12Okay? Otherwise, we might reject the
7:15strategy because the index used to be
7:17cheaper. We can we need to remember and
7:19use some common sense. We're uh
7:21deploying the strategy in an era where
7:24NQ is trading at 29k and it's only
7:27getting higher and higher and higher.
7:29So, the spread is not going to eat a lot
7:30of the strategy. All right. Now, let's
7:32look at the results. Phase one is the
7:34literal baseline. We follow the rules
7:36from the book as faithfully as possible.
7:38And the only part the book does not
7:40define mechanically is the exit or the
7:43TP.
7:44So, profit management was left mainly to
7:46traders judgement. We tested several
7:48exits, fixed targets at 2R, 3R, and 4R,
7:51and a trailing stop, and also a
7:53time-based exit.
7:55Um R simply represent amount risked on
7:58one trade. So, if we risk $100, then 1R
8:01is $100, 3R is 300, that's it.
8:05Now, two versions looked promising
8:07during development, closing after two
8:09session, so the time-based exit, and
8:12using a 3R target with a maximum hold of
8:15five sessions.
8:16From 2012 to 2017, both produced profit
8:20factor above 1.1.7.
8:23In simple terms, they made more than
8:25$1.7 for every $1 they lost. Okay, that
8:29sounds good, but the strategy didn't
8:31trade often. So, the frequency was too
8:34low, and each version made only around
8:3624R across five years. That's not good.
8:41And this is why one attractive profit
8:44factor number is not is never enough.
8:46Now, when we move to validation, both
8:49versions failed, as you can see. Uh the
8:51two session version dropped to 0.23 uh
8:54profit factor, a disaster. The 3R
8:57version dropped to 0.53, better, but
9:01clearly not profitable.
9:03But in the later periods, the two
9:04sessions version uh remained
9:06unprofitable, as you can see.
9:08And the 3R version actually recovered.
9:11But this strategy is ac- is obviously
9:13not tradeable because of the low
9:14frequency, and it actually uh went under
9:18for three whole years in the validation
9:20period from 2018 to 2020. However, we
9:23found another version that looked more
9:25promising. Uh instead of searching for a
9:27new 20-session extreme, we only used a
9:3010-session lookback.
9:33And it also removed the rule requiring
9:36the previous extreme to be at least four
9:38session old. Okay? Um so, we
9:41uh
9:42removed some strict guidelines, all
9:44right? And its results were actually
9:46stronger. So, it becomes our new
9:49baseline strategy for phase two. A
9:51descendant of the baseline.
9:54So, this is where Strategy Architect
9:56starts looking deeper. Okay? Instead of
9:59randomly changing indicators, it asks
10:01questions about what actually happens
10:03during the trade.
10:05As you can see, these are a couple of
10:06the questions it asks. You can go
10:08through them one by one. But basically,
10:10each idea must have a logical reason
10:13behind it. And we also have one major
10:15limitation, which is the strategy
10:17already has a low number of trades. That
10:19means we cannot keep adding filters. We
10:21cannot
10:22um
10:23because every filter is going to remove
10:25more setups, and eventually we would be
10:27left with five trades a year. At that
10:29point, one lucky trade could make the
10:31entire strategy
10:32look better than it really is. So,
10:35several ideas that uh Strategy Architect
10:37proposed looked excellent during
10:40development. One version added half a
10:43position only after the original trade
10:45had proven itself, so it's in profit.
10:48The 4-hour version looked even stronger,
10:50and combining the daily and 4-hour
10:53version also produced impressive
10:54results. But development is the easy
10:57part. The real question is whether those
10:59improvements
11:01uh survive later data. And most of them
11:03did not. Only one version consistently
11:06improved the parent, which is the half
11:08bank uh plus runner. And the idea behind
11:11this is very simple, uh the Strategy
11:13Architect came up with.
11:15When the trade reaches its normal exit
11:18on day 10, okay, we close half of the
11:21position, and we bank part of the
11:22profit.
11:24And the other half stays open for 10
11:27more sessions.
11:30So, basically, we are riding the trend,
11:32okay? Giving a strong move more time to
11:34continue. And NQ tends to does that.
11:38It's um a trending to the upside market.
11:42So, we're not adding more risk. We are
11:43paying ourselves with half of the
11:45position, all right? This version
11:47improved the parent during development,
11:48validation, and the final period. So,
11:51everywhere, the whole data set. But,
11:53there was still one problem we had not
11:55properly tested, which is overnight
11:57financing. Our enhanced version had
11:59become a swing trading strategy, and
12:01many trades stayed open for days or even
12:03weeks.
12:05And for the type of accounts we're
12:06targeting, that creates a big problem.
12:09Most futures prop firms require uh
12:11positions to be closed before the end of
12:12the day. So, we can't use them. And CFD
12:15prop firms allows us to hold positions
12:18overnight, but they charge a swap fee.
12:22And think of uh a swap fee is uh as a
12:26rent for keeping the position open,
12:28okay? And the average hold of this
12:30strategy or this version can be can go
12:32all the way to 15 days.
12:34Sometimes 30 days if we're going to
12:36catch a way bigger trend, you know? So,
12:39after swaps, nearly all of that profit
12:41disappeared, leaving the strategy around
12:43break even.
12:44And the only way that you can use this
12:45strategy is open a futures account, a
12:48personal futures account that does not
12:50have swaps. So, it's not something that
12:52everyone can use, all right? So, we
12:54found a profitable signal, but we did
12:56not have the right product or account
12:58structure to actually trade it.
13:01Now, why did the literal version from
13:03the book fail, which is the baseline?
13:05One possible and the most possible
13:07explanation is edge decay.
13:10An edge is simply a repeatable
13:12advantage. But but when an idea becomes
13:15widely known and more traders be begin
13:18exploiting it, that advantage can become
13:20smaller, okay?
13:22That is only one possibility. There is
13:24other possibilities, but that's a
13:26highest probable um
13:28ideal.
13:30And also research has shown that many
13:32market anomalies becomes weaker after
13:34they are published.
13:36And those are usually hidden inside
13:38academic uh papers that that very few
13:41people read. Now imagine a famous book
13:43containing simple strategies
13:45that millions of traders can understand
13:47and use.
13:49So if edge decay is present, the effect
13:51could be even stronger. All right? So
13:56what about the other strategies in this
13:57book? I haven't tested all of them, so I
14:00cannot honestly say they no longer work,
14:02but I also would not copy the rules
14:04directly and assume that
14:07they will produce the same results
14:08today. I highly doubt that.
14:10The real value of the book, in my
14:12opinion, is understanding the ideas
14:14behind the setups, how these uh traders
14:16identified failed breakouts, exhaustion,
14:19reversals, and change in market
14:20behavior. You can take those ideas,
14:22maybe add them as a part to your
14:24strategies. So yes, I still think the
14:27book is worth reading,
14:28but not as a collection of strategies to
14:31copy blindly, but um as a way to
14:34understand how legendary traders
14:36thought about the market back then.
14:39Um that's all for the video, guys. I
14:41hope uh you enjoyed it. If you want the
14:43detailed report, as always, leave a
14:45comment below,
14:47and make sure you are subscribed, and
14:49I'll send it to you. That's it.
14:51Now happy trading, and good luck.

About this transcript

This page contains the full transcript of I Backtested the Original Turtle Soup Strategy for 15 Years (ICT's Mentor) by Strange ICT, generated from the public captions YouTube serves with the video. The transcript has 2,495 words across 390 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.