I Backtested 4,200 Variation of Candle Range Theory CRT — Does It Actually Work? — Transcript

Full transcript

1. 0:00Hello guys and welcome back to a new
2. 0:02video. Today we're going to put the
3. 0:04candle range theory under the scope.
4. 0:07Probably the most hyped strategy on
5. 0:09social media right now and we're going
6. 0:11to find out if it's actually a real edge
7. 0:14or just garbage with heavy marketing
8. 0:16like the rest. In this video I tested
9. 0:194200 different variation of the strategy
10. 0:22and what I found out is going to
11. 0:23surprise a lot of people. So stick till
12. 0:26the end. But before we get into it, make
13. 0:28sure to check the current discount of
14. 0:30Mox is running. One of the most reliable
15. 0:32futures prop firms out there and they
16. 0:34just added two new paths on top of the
17. 0:37classic evaluation. So they added the
18. 0:39fast track, one-time fee, straight to
19. 0:42funded. So there is no evaluation. And
20. 0:44also they added the momentum path, a
21. 0:47one-time evaluation fee with no monthly
22. 0:50subscriptions and no activation fee once
23. 0:53you got funded. So make sure to check
24. 0:55them out. Use the coupon code stranger
25. 0:56or stranger 25. And now let's get into
26. 1:00the video. [music]
27. 1:01So as you can see on the screen,
28. 1:02everything I'm about to show you, I put
29. 1:04it into one big [music] report, okay? If
30. 1:07you want to read it in details or you
31. 1:09want to try different variations and
32. 1:11ideas yourself, all you have to do is to
33. 1:13leave a comment down below and I'll send
34. 1:16it to you, all right? So here's the
35. 1:18setup that we're going to do. I choose
36. 1:19um
37. 1:20the NQ futures. The data set runs from
38. 1:232021 all the way to 2026. So that's 5
39. 1:27and 1/2 years.
40. 1:28And um I got it from Data Bento. So it's
41. 1:32real high quality data.
42. 1:34And we split it into three parts. So the
43. 1:37first one is the in-sample from 2021 all
44. 1:39the way to 2024.
45. 1:41So this is where we do all our
46. 1:43exploring, finding variations,
47. 1:45optimizing and so on.
48. 1:48This is the playground. And [music]
49. 1:50the second part is the validation. So
50. 1:52that's the whole of 2025. The survivors
51. 1:55from the in-sample get their chance to
52. 1:57be tested here. All right? And finally,
53. 2:00the out-of-sample, which is 2026, quite
54. 2:04a small out-of-sample, but whatever
55. 2:07survives has to pass this, okay? And it
56. 2:10gets [music] only one shot. All right.
57. 2:13So, what is the candle range theory?
58. 2:17At a high level, the engine works like
59. 2:19this, okay? So, you take three candles.
60. 2:22Let's say we are in a downtrend. So, you
61. 2:24take a candle and treat its high and low
62. 2:27as a box or range.
63. 2:30So, when the next candle takes the high,
64. 2:32but fails to close above it,
65. 2:34the theory says it grabbed the stops and
66. 2:38>> [music]
67. 2:38>> closed back inside. That's called a
68. 2:40sweep, all right? And then, it's
69. 2:42supposed to reverse and run to the other
70. 2:44end of the original candle, okay?
71. 2:47So, once it happens, you go short and
72. 2:49target the other end and beyond.
73. 2:52That's the underlying engine. Of course,
74. 2:54there's more to it, and we'll get into
75. 2:55that later on. But, this isn't a new
76. 2:59idea. It is inspired by ICT power of
77. 3:01three and turtle soup, and both of those
78. 3:04are also repackaged from Wyckoff and the
79. 3:07swing trading methodology in the book
80. 3:09Street Smarts. So, I'm going to leave
81. 3:11the the links in the comment
82. 3:13to verify by yourself. And by the way,
83. 3:16if there is another famous strategy that
84. 3:17you want me to put it through the exact
85. 3:20pipeline
86. 3:21and the back testing engine, drop it in
87. 3:23the comments and I'll run the full back
88. 3:25test and see if it actually works or
89. 3:28not. I'll be happily to share the
90. 3:30results, including the exact variation
91. 3:32that actually made money, all right? So,
92. 3:34just tell me
93. 3:35down below.
94. 3:36Now, you might say that some strategies
95. 3:39or traders rely on discretion and that
96. 3:42you can't make every concept mechanical.
97. 3:45You can't back test it. And I agree to a
98. 3:47point, but this is a big misconception
99. 3:50in trading, all right? What most people
100. 3:52don't don't understand is that a real
101. 3:54institutional level trader
102. 3:56is not relying on feel or discretion
103. 3:59alone. Yes, there is discretion, but
104. 4:01underneath it, they have a solid
105. 4:03mechanics, concepts, and patterns in
106. 4:05price that wins more than they lose.
107. 4:09Okay? So, the underlying engine is
108. 4:11mechanical, and when they add
109. 4:13discretionary on top of that, the
110. 4:15results get even better.
111. 4:17In other word, discretion [music]
112. 4:19sits on top of a working mechanical
113. 4:21edge. Okay? So, discretion alone cannot
114. 4:24create an edge out of thin air. It needs
115. 4:27the mechanics underneath. Even if you
116. 4:30are watching me and you are a
117. 4:31discretionary trader, you must admit
118. 4:34that you have price action patterns,
119. 4:37um time frame alignments, higher time
120. 4:39frame analysis that you do
121. 4:42in order to make an idea of where the
122. 4:44market is going. So, you are a
123. 4:46discretionary trader, but you have a
124. 4:48mechanical um
125. 4:50engine underneath. Okay? So, that's my
126. 4:52point. So, in this video, we're
127. 4:55answering does this candle range theory
128. 4:57mechanical core have any edge [music] at
129. 5:00all before any human discretion to add.
130. 5:04All right?
131. 5:05So, let's talk about the variations that
132. 5:07I created. So, I wanted to give CRT
133. 5:10every fair chance, so we built 40 to 100
134. 5:13versions or variation. First of all, the
135. 5:16time frame combination, okay? As you can
136. 5:18see, six different ones.
137. 5:20Then, the ways of finding a sweep or
138. 5:23defining a sweep.
139. 5:24>> [music]
140. 5:25>> When a candle takes a high, there is
141. 5:27more than one way to call it or to
142. 5:29define it. So, I tested 10 of them. As
143. 5:32you can see, [music] these are all on
144. 5:33the screen. Same for stop loss
145. 5:35placement. Uh stop loss placement, guys,
146. 5:37can be a real edge on its own. A lot of
147. 5:39people think that stop doesn't matter,
148. 5:42but
149. 5:43the right placement, like [music] um
150. 5:45either it's tight or wide stop loss or
151. 5:48ATR based stop loss or something more
152. 5:50advanced like a trailing stop loss can
153. 5:52literally take a strategy from
154. 5:54unprofitable to profitable. Okay? So, in
155. 5:57this
156. 5:58field we tested seven. And the same
157. 6:00thing for take profit. So, to put it all
158. 6:03together and to multiply everything we
159. 6:05have 42 separate strategy combination
160. 6:09variation call it what you want.
161. 6:11And we run every single one through the
162. 6:14back testing engine. Okay?
163. 6:17And as you can see these are the
164. 6:19results. I want you to pay close
165. 6:22attention to this graph right here.
166. 6:24Okay? Uh on the x-axis is the profit
167. 6:27factor. Okay? Remember profit factor
168. 6:29just means for every dollar you lose how
169. 6:32much do you win? Okay? So, a profit
170. 6:34factor of 1.5 means for every dollar
171. 6:37that you lost you make 1.5 and that's
172. 6:40good. Okay? That's something that you
173. 6:41want to have.
174. 6:43So, above one is profitable, below one
175. 6:45is unprofitable,
176. 6:47but you really want something above 1.3.
177. 6:50Okay? At least 1.2.
178. 6:53Otherwise, it's just a thin edge that
179. 6:55you cannot rely on.
180. 6:56>> [music]
181. 6:57>> Now, I want you to look at the
182. 6:58distribution. All right? About 10% had a
183. 7:02really terrible profit factor around 0.5
184. 7:05and below. And another 10% had a really
185. 7:08good one. So, from 1.8 all the way to
186. 7:11three.
187. 7:12But, here's the important part that you
188. 7:14should focus on. 80% of those 4200
189. 7:18variation landed around 0.92 profit
190. 7:22factor.
191. 7:24Okay? And this matters a lot. All right?
192. 7:26Because if you can actually
193. 7:28have a real phenomenon in price
194. 7:31when you test all these variations what
195. 7:34you want to see is that most of them
196. 7:36lean to the profitable side above one.
197. 7:39Okay? And that's not what's happening
198. 7:40here. So, this is a good sign. You might
199. 7:44um point [music] to the 10% with a good
200. 7:46profit factor.
201. 7:48Uh and that's 10% uh across 42
202. 7:50variations. That's like um
203. 7:52>> [clears throat]
204. 7:53>> 400 variations that really had
205. 7:55a good profit factor.
206. 7:58But, it's not enough. 10% is not good.
207. 8:02Okay? It might be most of them can be
208. 8:04just overfit to noise. But, remember
209. 8:07that we're still on the first split.
210. 8:09Okay? So, this is the in sample. And now
211. 8:11what we're going to do is we're going to
212. 8:13take [music] those survived, the 10%
213. 8:15that survived, um the ones with the
214. 8:18decent profit factor, and we're going to
215. 8:19test them on the rest of the splits. So,
216. 8:22we have the the validation of 2025, and
217. 8:25finally the out of sample. Okay?
218. 8:27But, before that, we're going to run
219. 8:29them through something called the
220. 8:31placebo test. This is a well-known test
221. 8:34in algorithmic strategy development, and
222. 8:36it goes like this. Okay? I want you to
223. 8:39listen to this one.
224. 8:42So, for each variation that passed the
225. 8:45in sample, okay? We're going to build a
226. 8:48dump twin.
227. 8:50It's going to take the same exact trades
228. 8:52at the same times, same same sizes,
229. 8:56okay? And the same stop loss and take
230. 8:58profit.
231. 9:00The only thing that changes is
232. 9:02direction. So, instead of using the CRT
233. 9:05signature, which is sweeping the
234. 9:07previous high and then closing below,
235. 9:09therefore, we need to go short. No,
236. 9:12we're going to throw that away, and when
237. 9:14it's time to take the trade, we're going
238. 9:15to flip a coin.
239. 9:17If it is tails, we're going to go short.
240. 9:20>> [music]
241. 9:20>> If it heads, we're going to go long. And
242. 9:22that's it. That's only thing that we're
243. 9:23going to change. Now,
244. 9:25when you think about it for a second, um
245. 9:28because I know it sounds pointless,
246. 9:30because if the candle range duty is
247. 9:32actually reading the market,
248. 9:34if that sweep really holding an edge on
249. 9:38direction if that formation of price or
250. 9:41candles then the real CRT variation
251. 9:45should crush the coin flip twin. It's
252. 9:47not even a fair fight, all right?
253. 9:52It's the smart one against the random
254. 9:54one. But when I ran it guys, um they
255. 9:57basically tied. A lot of the variations
256. 10:01that passed the in sample,
257. 10:03okay?
258. 10:05They failed the placebo test. Their dumb
259. 10:08twin made almost the same money, almost
260. 10:11the same profit factor, okay? So, that
261. 10:14will tell you that it wasn't the sweep,
262. 10:16it wasn't that candle formation, okay?
263. 10:18[music]
264. 10:19Giving those winners an edge. It was
265. 10:21placing the right stop and the right
266. 10:23target, okay? And a lot of people don't
267. 10:26like running this test and they feel
268. 10:28like it's pointless, okay? Because
269. 10:30[music] they don't like the result, but
270. 10:31it is guys. It tells you whether your
271. 10:33strategy is actually based on a real
272. 10:36thing, okay?
273. 10:37If you want more detailed video about
274. 10:40the placebo test and other well-known
275. 10:41tests that you should run on your
276. 10:43strategies, okay? When you are
277. 10:44developing strategies, tell me in the
278. 10:46comments and I'll do
279. 10:48uh one with more examples.
280. 10:50But
281. 10:51uh in summary, the placebo test
282. 10:53eliminated a large chunk of the
283. 10:55variations that had passed on the profit
284. 10:57factor in the in sample, okay? It turned
285. 11:00out that a lot of them it wasn't [music]
286. 11:02because of that particular signature
287. 11:05that brought in price, okay?
288. 11:07So, after the placebo test cut um
289. 11:11couple of variations,
290. 11:12we looked at the equity curves at two
291. 11:15variations that really stood out because
292. 11:18one of them made almost 743
293. 11:21R. That's insane, okay? That's a really
294. 11:24good number. And another it made even
295. 11:27better.
296. 11:28It made 1,121
297. 11:31R,
298. 11:32Okay?
299. 11:33>> [music]
300. 11:33>> Now, both
301. 11:35had a strong profit factor and when I
302. 11:38looked at their equity curves,
303. 11:41it was extremely disappointing.
304. 11:44As you can see, these taken extremely
305. 11:46low number of trades, as you can see on
306. 11:48the chart, the frequency is really not
307. 11:51there and one of them literally took a
308. 11:53single trade in the span of 2 years.
309. 11:56So, the conditions were really hard to
310. 12:00replicate
311. 12:02um every day or every week. And as you
312. 12:05can see, they made almost all their
313. 12:06money in a couple of lucky windows
314. 12:09with deep drawdown in between. So, in
315. 12:11other words, they are useless. We cannot
316. 12:13use them.
317. 12:15And the rest of the variations that beat
318. 12:17the placebo in sample and had a decent
319. 12:20profit factor and healthy equity curve
320. 12:23and healthy frequency, so they passed
321. 12:26all the tests and so on,
322. 12:29when they when we tested them on the
323. 12:30validation,
324. 12:32which is 2025 and the out of sample in
325. 12:352026,
326. 12:36that placebo
327. 12:38caught up.
328. 12:40Those coin flips
329. 12:42or dumb twins that we created in the in
330. 12:44sample, they recovered in the validation
331. 12:47sample and the out of sample. Okay? Same
332. 12:50profit factor, same overall profit,
333. 12:52[music] which means even for the
334. 12:54survivors, okay,
335. 12:57that passed all the tests, the sweep,
336. 12:59the actual candle range theory engine
337. 13:01was not the one producing those numbers
338. 13:04because a dumb twin that randomized
339. 13:06direction, okay, made pretty much the
340. 13:09same money. So, it tells you that the
341. 13:11money
342. 13:12it came from the structure,
343. 13:14the overall structure, not that
344. 13:17signature in price. So, does the candle
345. 13:19range theory have a real mechanical edge
346. 13:22that you can trade
347. 13:23without the discretion that they are
348. 13:25trying to sell you?
349. 13:26>> [music]
350. 13:27>> The kind that you spend years of your
351. 13:29life chasing based on 42 hundred for
352. 13:32Asian that we tested, the placebo test,
353. 13:35and the sealed
354. 13:37validation sample, and the out of sample
355. 13:39data, the answer is no. It doesn't have
356. 13:43an underlying engine that drives the
357. 13:45edge in your favor. Okay? The typical
358. 13:48the typical CRT version loses money. A
359. 13:51profit factor of 0.92. You saw the
360. 13:53distribution. A real edge isn't that
361. 13:55fragile. And let me prove it to you.
362. 13:57Okay? I'm going to show you one. This is
363. 13:59an edge that I'm testing on the Forex
364. 14:01side
365. 14:03or the commodity side on gold. I want
366. 14:05you to look at distribution of the
367. 14:08variations and compare to CRT.
368. 14:11Most of the variations lean [music] to
369. 14:13the right into a good profit factor.
370. 14:15That's what a real edge looks like.
371. 14:18Okay? CRT is in the middle. It has
372. 14:21nothing to offer. All right? And you can
373. 14:23[music] actually add discretion to this
374. 14:25gold strategy and make it even better.
375. 14:29So you're adding discretion into
376. 14:31something that actually works.
377. 14:33All right? Now again,
378. 14:35if you want me to test other things, if
379. 14:37you want
380. 14:38if you maybe notice that I missed some
381. 14:41things, I didn't give it the fair chance
382. 14:44to actually work, maybe I missed couple
383. 14:46of mechanics, I'm happy to test this
384. 14:49again in another video. Just leave them
385. 14:52in the comments and I'll see. Okay?
386. 14:54We're going to let the data decide. But
387. 14:57for now, this is the end of this video.
388. 14:58If you have any other questions, leave
389. 15:00them down in the comments and I'll
390. 15:02answer you as soon as I can. Now happy
391. 15:04trading and good luck.