Building a Profitable Strategy w/ Claude - Day 20: THIS GONNA BLOW YOUR MIND! — Transcript

Full transcript

1. 0:00Hello guys, and welcome back to a new
2. 0:01video in the series of me trying to
3. 0:03build a profitable trading algorithm
4. 0:05using AI.
5. 0:07>> [music]
6. 0:07>> In today's video, I'm going to share
7. 0:09with you something that's going to blow
8. 0:11your mind away. By the end of it, the
9. 0:13way you think about using AI to find
10. 0:15trading strategies is going to
11. 0:17completely change.
12. 0:19Also, we're going to address and clear
13. 0:21one of the biggest misconceptions people
14. 0:24[music] have about this whole thing. So,
15. 0:26here's where we are. In the last video,
16. 0:28I showed you two strategies, one trend
17. 0:30following and one mean reversion. But,
18. 0:33here's the thing. Those were just two
19. 0:35ideas, two hypotheses out of literally
20. 0:39hundreds we could test.
21. 0:41>> [music]
22. 0:41>> And on the backtest, both of them failed
23. 0:43as I told you. I didn't show you the
24. 0:46metrics and the results, but they both
25. 0:48failed.
26. 0:49Now, after that video, a lot of people
27. 0:52asked me in my DMs about how does the
28. 0:56backtest actually work? All right?
29. 0:58Because we get how the AI comes up with
30. 1:01the hypothesis I showed you, it reads
31. 1:03the research online, [music]
32. 1:06scans thousands of papers. We get how
33. 1:08the data works, all right? We pulled it
34. 1:11from a vendor like Data Bento, or you
35. 1:13can get from any kind of source. But,
36. 1:15the backtest itself, the conditions, the
37. 1:18parameters, the filters, the indicators,
38. 1:22does it optimize the strategy while it
39. 1:24runs? Does it go and find other
40. 1:26variations of the strategies? I want you
41. 1:28to focus on the word variations because
42. 1:31this is exactly where this is going to
43. 1:33get crazy.
44. 1:34Today, I'm going to take you under the
45. 1:36hood, okay? And show you exactly what
46. 1:38happens when we run a backtest.
47. 1:41>> [music]
48. 1:42>> So, not just on the strategy, but on any
49. 1:44strategy. And I promise you, this is
50. 1:47going to open your eyes. All right? Now,
51. 1:49before we start, make sure to check the
52. 1:51current discount Balonkus running, one
53. 1:53of the best and most reliable futures
54. 1:55pro firms out there. You can use a
55. 1:57coupon code STRANGER or just click on
56. 2:00the link down below.
57. 2:01Now, let's get started.
58. 2:04So, we're going to start with this
59. 2:06picture that's going to
60. 2:08explain every single thing about how the
61. 2:10back test works. Okay?
62. 2:12So, this tree right here, we're going to
63. 2:14start from the root. All right?
64. 2:17We're going to start from a strategy
65. 2:18idea, a hypothesis. This is how AI goes
66. 2:22to, like I showed you, how it got that
67. 2:24PDF. It's going to go online and it's
68. 2:27going to choose for us a hypothesis that
69. 2:30we need to test. Okay? This is just a
70. 2:33simple one, all right? A strategy idea.
71. 2:36This strategy idea is based on the
72. 2:37moving averages. All right? So, a trend
73. 2:40following strategy that uses moving
74. 2:43averages.
75. 2:44Simple. Everybody knows what the moving
76. 2:47average is.
77. 2:48So, we're going to start from here, the
78. 2:50root.
79. 2:51Now,
80. 2:51>> [music]
81. 2:51>> we need to understand that moving
82. 2:53averages, we have two types. Okay? I
83. 2:55want you to look at the left. This is
84. 2:57the first layer and basically these
85. 3:00right here are just conditions for the
86. 3:01strategy. Okay? Let's start with layer
87. 3:04one.
88. 3:05For the averaging, there is two types.
89. 3:07There is the simple moving average and
90. 3:09there is the exponential moving average.
91. 3:11Now, instead of guessing which one is
92. 3:12better for this strategy, we're going to
93. 3:15test both of them. So, from the root,
94. 3:18we're going to have the simple moving
95. 3:20average and we're going to have the
96. 3:21exponential moving average. We're going
97. 3:23to test both and see which one is
98. 3:24better. Okay?
99. 3:27Now, after that, we need fast moving
100. 3:30average.
101. 3:31Okay? This going to catch like
102. 3:34um recent break of structures or an
103. 3:38early um
104. 3:39So, let me show you on the chart. As you
105. 3:41can see, this red moving average right
106. 3:44here, this is the exponential moving
107. 3:45average, it has a length or period of
108. 3:48nine. Okay? So, as you can see, it's
109. 3:50going to catch those um break of
110. 3:53structures like here, like shifts in
111. 3:55trends.
112. 3:56E- even if they they fail, for example,
113. 3:59this one didn't fail because you can see
114. 4:01that it did provide some uh movement
115. 4:04after this uh up close candle here. Of
116. 4:07course, we're not going to use this one
117. 4:10fast moving exponential moving average
118. 4:12to catch a big shift in trend. It
119. 4:13doesn't work like that. We're going to
120. 4:15see how we can catch this
121. 4:17uh using this strategy alongside with
122. 4:19other conditions. So, we're going to
123. 4:21need this, all right? So, we need a fast
124. 4:24moving average. Now, fast moving
125. 4:26averages,
126. 4:27there is different periods or different
127. 4:30parameters. [music]
128. 4:31There is a 2 3 5 8 9 and 13. We can go
129. 4:35all the way to 10 and 12, but there is
130. 4:37not big difference between. So, it's
131. 4:39better to keep uh some distance between
132. 4:41the parameters.
133. 4:43And we don't know which one is better.
134. 4:45Which one is going to give us a strategy
135. 4:47that works. So, we're going to test both
136. 4:49of them.
137. 4:50So, on the SMA is we're going to test
138. 4:52all of these.
139. 4:53And for the EMA, which is which is here,
140. 4:56we're going to test all these periods,
141. 4:58as you can see. So, five here we have
142. 5:00five on each side, okay?
143. 5:04Now, for this strategy, we also need a
144. 5:06slow moving average. A slow moving
145. 5:08average basically the same thing as this
146. 5:11one, but we're going to use higher
147. 5:13values, 20 30 45 60 and [music] 90. So,
148. 5:17as you can see on the chart, for
149. 5:18example, this one is the EMA 50. And you
150. 5:22can see why it is a slow because it
151. 5:24doesn't react and change to
152. 5:27uh these kind of uh shifts in structure.
153. 5:30Okay, it catches or it gives you the big
154. 5:32the bigger picture, all right? So, you
155. 5:34can see
156. 5:36from here to here we are on we are in a
157. 5:38downtrend, okay? Here we were in an
158. 5:41uptrend, and then price broke below the
159. 5:44EMA 50 and also the nine. so both the
160. 5:48slow and and the fast moving average and
161. 5:51we had a nice run to the downside and
162. 5:53here you can use this one as directional
163. 5:55bias and only look for shorts. Okay? So
164. 5:58basically this
165. 6:00um temporarily shift in structure,
166. 6:01you're not going to take it seriously
167. 6:03because we're still overall down
168. 6:06um in a downtrend based on the slow
169. 6:09moving average. Okay? This is how it
170. 6:10looks like in the chart.
171. 6:12Now,
172. 6:14after that,
173. 6:16we're going to need a filter or a
174. 6:17directional bias. So think of it like
175. 6:20the higher time frame. For that we're
176. 6:21going to use either none, we're not
177. 6:24we're not going to use that,
178. 6:25or we're going to use the moving average
179. 6:27of 100,
180. 6:29200, and the ADX above 20 or ADX above
181. 6:3325.
182. 6:35Now, here's the thing.
183. 6:36For example, a strategy
184. 6:39it's going to use the EMA.
185. 6:41Okay?
186. 6:43For the fast EMA it's going to use the
187. 6:44five.
188. 6:46For the slow
189. 6:47EMA it's going to use the 60 and for the
190. 6:50filter it's going to use the moving
191. 6:52average of 200. This is one strategy.
192. 6:55Another one it's going to use the SMA
193. 6:57with three, with 20, and then ADX.
194. 7:00Another one is going to use
195. 7:02SMA of eight. So these are all
196. 7:05combination. Okay? This is why we have
197. 7:07multiplier here.
198. 7:08All [music]
199. 7:09possible combinations of a strategy. So
200. 7:12based on just these first four layers,
201. 7:15we have like two times five times five
202. 7:18times five. Like we have 250 [music]
203. 7:23possible and different combination,
204. 7:26different strategies with different
205. 7:28parameters.
206. 7:30Moving on to the stop, stop loss, which
207. 7:33is based on the ATR. ATR is very simple.
208. 7:36Let me show you in the chart. So we can
209. 7:39add the ATR from here.
210. 7:43So, basically, suppose that we are here,
211. 7:46okay? We want to take a trade
212. 7:49here.
213. 7:52All right? We're going to take a short.
214. 7:54Or maybe take a long. Let's take a long.
215. 7:57So, [music] how does this strategy uses
216. 7:59the stop loss?
217. 8:01Okay? It's going to use this ATR, the
218. 8:04average true range. I showed you what
219. 8:06the ATR is. It's going to It's going to
220. 8:07basically take the last 14 candles right
221. 8:10here. We can change the volume as well,
222. 8:12but we're going to use the 14. It's
223. 8:14going to calculate or it's going to take
224. 8:17the last 14 candles. It's going to
225. 8:19calculate their sizes, sum them, okay?
226. 8:22So, add them all together, and then
227. 8:24dividing my by 14. And it's going to
228. 8:26give us this volume right here. For
229. 8:28example, on this, it is
230. 8:29>> [music]
231. 8:29>> 76. Let's say 75. So, um if we go back
232. 8:34to the picture,
233. 8:36if we're going to use 1.5,
234. 8:39or let's use the stop of two two ATR,
235. 8:43all right?
236. 8:44That's going to give us basically
237. 8:45hundreds and [music]
238. 8:4850.
239. 8:49So, the stop loss is going to be right
240. 8:51here. So, the stop loss is going to
241. 8:53change based on the ATR. So, if the ATR
242. 8:56is like 100, as you can see here,
243. 8:58because uh as you can see, the candles
244. 9:01are bigger and the volatility is higher
245. 9:03than this, it's going to be 100. For
246. 9:05example, if we take trade long
247. 9:08right here,
248. 9:10so, the the ATR is going to be 200 like
249. 9:12this. As you can see, it's based on the
250. 9:14volatility. Now, we don't know which ATR
251. 9:18uh stop is stop loss is better, so we
252. 9:20need to test different volumes. Okay?
253. 9:22So, we're going to use tight stop loss,
254. 9:25only 1.5 times the ATR.
255. 9:27Uh we're going to use moderate stop
256. 9:29loss. We're going to use wide stop
257. 9:31losses. Okay? Times six.
258. 9:33All right?
259. 9:34So, we know which stop loss is better.
260. 9:37Same thing for the trailing, we can use
261. 9:39ATR, so trailing the stop loss, and also
262. 9:41the max hold to the TP, and the
263. 9:43direction long and short. Okay, I spoke
264. 9:45a lot,
265. 9:47I know,
266. 9:48but
267. 9:49here's what's going to blow your mind
268. 9:50away.
269. 9:52Based on these parameters,
270. 9:54we we're going to have 15,000
271. 9:58strategy combination.
272. 10:01Imagine this.
273. 10:01>> [music]
274. 10:02>> 15,000
275. 10:04Each leaf right here, as you can see,
276. 10:07this is a leaf,
277. 10:08is going to have one unique [music]
278. 10:10strategy combination. Unique.
279. 10:14Different, completely different. So,
280. 10:1515,000
281. 10:17strategy or variation that came out of
282. 10:19this one simple idea.
283. 10:22And this 15,000 strategy,
284. 10:25okay?
285. 10:27We're going to use it
286. 10:29on nine different markets, as you can
287. 10:31see here. We have the US 100, US 30, we
288. 10:34have gold, we have Brent,
289. 10:36uh we have USD uh some forex pairs.
290. 10:39We're going to test it, so that's times
291. 10:41nine.
292. 10:43Also, we're going to use it on different
293. 10:44time frames, or five different time
294. 10:46frames. I didn't include this. We're
295. 10:47going to use it on the five-minute, the
296. 10:4915-minute, the 30-minute, the one-hour,
297. 10:51and the four-hour. So, times four.
298. 10:55That's going to give us 540,000
299. 11:00back test.
300. 11:02That means 63.6
301. 11:04billion [music] candle evolution.
302. 11:07And that's going to be running
303. 11:09simultaneously,
304. 11:10okay?
305. 11:11On seven CPU cores. In my laptop, I have
306. 11:15eight, okay? You might have way more.
307. 11:19So, basically,
308. 11:21it's going to run all this back test.
309. 11:24It's going to run
310. 11:25540 simultaneous back test.
311. 11:29Just imagine, guys.
312. 11:31It is literally insane.
313. 11:34And
314. 11:35it's going to take around
315. 11:3750 minutes.
316. 11:39If you have a better laptop, it's going
317. 11:41to take less. For me, here's how much it
318. 11:44took. So, as you can see here, this is a
319. 11:45picture that I took earlier.
320. 11:47Uh this is a sweep. The sweep is is
321. 11:50basically what's happening here,
322. 11:51sweeping across all this. This is the
323. 11:53back test. Okay? And as you can see, it
324. 11:57is first of all trying um
325. 11:59gold on the M15.
326. 12:02So, basically, gold, we're choosing
327. 12:04gold, and we're choosing M15.
328. 12:08Now, it's running 15,000 back tests.
329. 12:1015,000 strategy is running on the
330. 12:1415-minute time frame on gold.
331. 12:17This is what this represent.
332. 12:20And this
333. 12:22And out of those 15,000, only six
334. 12:26survived.
335. 12:27Only six strategies survived.
336. 12:30And the best the best profit factor was
337. 12:321.12. Now, here's the thing.
338. 12:351.12 profit factor and only six per
339. 12:37survived out of those 15,000, it's not
340. 12:40good.
341. 12:41Okay, let me tell you why.
342. 12:44You want to see something bigger,
343. 12:45something like this. It tells you that
344. 12:48that particular strategy on this
345. 12:51M15 on US 100 here,
346. 12:54it actually works.
347. 12:55Because a lot of them actually passed um
348. 13:00the current metrics like profit factor
349. 13:01and so on above one and couple of other
350. 13:04uh standards that I set. We can talk
351. 13:06about them later.
352. 13:08But it tells you that this strategy
353. 13:10right here is real on this one.
354. 13:13Okay, it also passed the cost and the
355. 13:17slippage and spread and so on. Okay? And
356. 13:20the best profit factor, the best
357. 13:22strategy, best uh leave like this one,
358. 13:26It has this 1.22
359. 13:29profit factor. Now, this is on M15.
360. 13:33Remember that on M15 is going to have
361. 13:35spreads and a higher cost.
362. 13:38Um I'm pretty sure that on the 1 hour
363. 13:41it's going to be even better, even
364. 13:42better profit factor and better metrics
365. 13:45overall. So, this is just to show you
366. 13:48how does this work. Okay? Now, imagine
367. 13:51trying to run this all alone. If we take
368. 13:53only one leaf out of 15,000 um
369. 13:58possible,
370. 13:59how much it takes that you can
371. 14:02to backtest it. And here's the thing,
372. 14:05I'm running I'm running it on for
373. 14:06example on this one this on Forex.
374. 14:09I have 16 years of data
375. 14:12from 2010 all the way to 2026
376. 14:17running on running at the same time. And
377. 14:20this one 40 minute and it was estimated
378. 14:24time to finish 70 minutes. So, it's
379. 14:26going to take you around two two and a
380. 14:28half hours
381. 14:30to three hours based on
382. 14:33the power of your laptop. Okay?
383. 14:36Or computer. So, as you can see here,
384. 14:38you don't find robust strategies by
385. 14:40guessing, you find them by exploring the
386. 14:43space. Okay?
387. 14:45This is how powerful backtesting with AI
388. 14:48[music] is.
389. 14:50Okay, guys? So, for example here here's
390. 14:52an active path. Here's an active leaf or
391. 14:54a
392. 14:55a single
393. 14:56um
394. 14:57strategy out of those 15,000 possible.
395. 15:01Choosing for layer one, we choose the
396. 15:04What is this? Uh the simple moving
397. 15:06average.
398. 15:07For the fast one,
399. 15:09we can choose the input of
400. 15:11uh five the period. For the slow one,
401. 15:14okay? I'm using uh these layers. Okay?
402. 15:1945 for the slow one. And for the
403. 15:21directional bias you're going to use the
404. 15:23moving average of 100.
405. 15:25For the stop loss, we can use three
406. 15:27times the ATR. For the trailing, we're
407. 15:29going to use four times ATR. And [music]
408. 15:32seven layer is um max hold two, okay?
409. 15:36And we're going to test long because of
410. 15:38course we're going to test both long and
411. 15:40short. And say for example in US 100, um
412. 15:44longs tend to be better, okay? Since it
413. 15:46is a trending um
414. 15:49instrument trending to upside. So this
415. 15:51is just one single path out of 15,100.
416. 15:56Okay? And this, ladies and gentlemen,
417. 15:59how powerful trading and finding
418. 16:02strategies with AI is. Now make sure to
419. 16:05like and subscribe the video so you
420. 16:06won't miss the next one. I want [music]
421. 16:08you to be active with me in this journey
422. 16:11not just what by watching but actually
423. 16:13uh trying it yourself, okay? So go to
424. 16:16the lab and
425. 16:18give your um Claude or ChatGPT or
426. 16:22whatever you're using
427. 16:24this picture, feed it this picture and
428. 16:26it's going to create the engines to
429. 16:28actually
430. 16:30uh sweep the space like this, okay?
431. 16:33Now, happy trading and good luck.