【AI手法完成】AIが作ったゴールド攻略法。その売買ルールと検証結果を公開 — Transcript

Full transcript

1. 0:00Hello. It's Reimu.
2. 0:01>> And Marisa.
3. 0:02>> Today, we’re using only Gold and a
4. 0:05moving average with a 35-40%win rate and
5. 0:07a 1: 3 risk-reward ratio. When
6. 0:10backtested on past charts, we found a
7. 0:11strategy that creates an upward profit
8. 0:13curve like this.
9. 0:15>> It's a beautiful upward slope.
10. 0:17>> Since all the rules are numerical,
11. 0:19anyone who uses it will enter and exit
12. 0:21at the exact same spots.
13. 0:23>> Anyone can do it, and it gets these
14. 0:25results? Looks like we’ve finally found
15. 0:27the Holy Grail.
16. 0:29I’d love to say that, but I need to be
17. 0:31honest with you first.
18. 0:32>> What is it?
19. 0:33>> Actually, I’ve never entered a trade
20. 0:35using this strategy even once.
21. 0:37>> Wait, not even once?
22. 0:39>> And it’s not even a strategy I came up
23. 0:41with; it was created by an AI.
24. 0:44>> That’s a sudden confession.
25. 0:46Then what was that graph from earlier?
26. 0:48>> That’s the result of backtesting past
27. 0:50charts using the rules the AI found. So
28. 0:52, you haven’t actually made any money
29. 0:54with this yourself, have you?
30. 0:56>> Ugh, you almost tricked me!
31. 0:58When I introduced a strategy before, I
32. 1:00backtested my own idea with AI, but this
33. 1:02time, I let the AI handle everything,
34. 1:04even finding the strategy.
35. 1:07And this is the graph that it came up
36. 1:08with.
37. 1:10>> AI can do things like that?
38. 1:12>> Yes, that’s exactly what makes it both
39. 1:14convenient and a bit scary.
40. 1:16So, this video is a strategy
41. 1:18introduction, but also a warning to
42. 1:20everyone.
43. 1:21>> I see.
44. 1:22>> I’ll explain that part in detail later.
45. 1:25First, I’ll explain the rules with an
46. 1:27image, then we’ll look at the actual
47. 1:29charts. Please watch until the end.
48. 1:31>> A must-see until the end.
49. 1:33Let’s get right into it.
50. 1:36In short, the strategy is to buy Gold on
51. 1:38the 1-hour chart during an uptrend,
52. 1:40waiting for a dip to the 20 EMA and
53. 1:42confirming it's starting to rise again.
54. 1:45The stop-loss is just below the low of
55. 1:46the dip. The take-profit is three times
56. 1:48the stop-loss range. We only buy; we
57. 1:50don’t use shorts.
58. 1:51>> It's quite simple.
59. 1:53From here, we’ll look at the specific
60. 1:55rules using a diagram. There are three
61. 1:57steps. The first is trend confirmation.
62. 2:00The second is the entry. The third is
63. 2:01stop-loss and take-profit.
64. 2:03>> I'm counting on you.
65. 2:04>> First, step one.
66. 2:06We confirm if it's a good market to buy
67. 2:08. There are four conditions. On the 1-
68. 2:10hour chart, the 20 EMA must be above the
69. 2:13200 EMA. The closing price must be above
70. 2:15the 200 EMA. The 200 EMA must be higher
71. 2:18than it was 20 bars ago.
72. 2:20Also, on the completed 4-hour chart, the
73. 2:2220 EMA must be above the 200 EMA and the
74. 2:25close above the 200 EMA. 1
75. 2:27>> So it's limited to when both the 1-hour
76. 2:29and 4-hour charts are pointing up?
77. 2:31>> Exactly, to avoid going against the
78. 2:33major trend. If even one condition is
79. 2:35missing, we do nothing with this
80. 2:37strategy. Next, step two: where to buy.
81. 2:40Even in an uptrend, prices don't rise in
82. 2:42a straight line; they often dip toward
83. 2:44the 20 EMA.
84. 2:46This strategy targets those pullbacks
85. 2:47instead of jumping in while it's rising
86. 2:49.
87. 2:50The condition is that the low has
88. 2:52reached the 20 EMA within the last 12
89. 2:54bars.
90. 2:55And that no candle closed below the 200
91. 2:57EMA during those 12 bars.
92. 2:5920
93. 3:00>> So it's a pullback that hit the EMA but
94. 3:02didn't break the 200 EMA?
95. 3:04>> That's right.
96. 3:05But we don't buy just because it dipped
97. 3:07. It must close as a bullish candle
98. 3:09above the 20 EMA, breaking above the
99. 3:11previous candle's high.
100. 3:13Once all three align, confirming it's
101. 3:15rising, we buy at the open of the next
102. 3:17candle.
103. 3:18>> To avoid buying while it's falling and
104. 3:19watching it drop further, right?
105. 3:21>> Yes.
106. 3:23And we only enter once per pullback.
107. 3:25Step three: the stop-loss is slightly
108. 3:27below the lowest price of the last 12
109. 3:29bars. Specifically, 0.01%below that
110. 3:32price.
111. 3:34Since the pullback low is clear, we can
112. 3:36keep the stop-loss range small.
113. 3:38Take-profit is at three times the
114. 3:40stop-loss range. For example, if the
115. 3:42stop-loss distance is $ 10, take-profit
116. 3:45is $ 30 above the entry.
117. 3:46>> Determining the take-profit level
118. 3:48mechanically by a multiplier is very
119. 3:50AI-like.
120. 3:50>> We don't move the stop to breakeven or
121. 3:52extend profit targets mid-trade.
122. 3:54The graph at the beginning was also
123. 3:56based on following these rules.
124. 3:58>> So, if the rules change, the results
125. 4:00will change too?
126. 4:01>> There are three cases where you should
127. 4:03skip, even if the conditions are met.
128. 4:05When the stop-loss range is smaller than
129. 4:070.3 times the average 1-hour candle
130. 4:09length.
131. 4:10That’s because it’s too tight and you'll
132. 4:12lose out to costs. Conversely, if it’s
133. 4:14more than 3 times larger, it’s too wide
134. 4:16, so skip that too. And finally, when
135. 4:18you already have a position open.
136. 4:20>> So there are times when you don't enter
137. 4:21even if the conditions are met.
138. 4:22>> That concludes the entry rules.
139. 4:25Now, let's take a look at an actual
140. 4:26chart.
141. 4:27>> I'm counting on you.
142. 4:29>> This is a 1-hour chart of Gold. First,
143. 4:31step one. On the 1-hour chart, the 20
144. 4:33EMA is above the 200 EMA, and the
145. 4:35closing price is also above the 200 EMA
146. 4:38. The 200 EMA has been rising for 20
147. 4:40periods, and it also met the conditions
148. 4:42on the confirmed 4-hour candle.
149. 4:44>> So step one is a go. Next, step two.
150. 4:48In the last 12 candles, the low dropped
151. 4:50to the 20 EMA, and no candles closed
152. 4:52below the 200 EMA during that time.
153. 4:55While waiting here, a bullish candle
154. 4:57closed above the 20 EMA, breaking the
155. 4:59high of the previous candle. So, we go
156. 5:01long at the open of the next candle.
157. 5:04Where are the stop-loss and take-profit
158. 5:06set?
159. 5:06>> The stop-loss is set slightly below the
160. 5:08lowest low of the last 12 candles, with
161. 5:11a 1: 3 risk-reward ratio.
162. 5:13>> After that, it climbed nicely for a win
163. 5:15.
164. 5:16>> If you check the conditions in order,
165. 5:18anyone can enter at the same spot.
166. 5:20Now, look at this one. This is another
167. 5:22scene where the conditions are met for a
168. 5:24long entry.
169. 5:25>> Going up again this time.
170. 5:26>> However, this one is a loss.
171. 5:28It hit the stop-loss.
172. 5:30>> So you can lose even when the conditions
173. 5:32are met?
174. 5:33Yes, this strategy has more losses, but
175. 5:35recovers with occasional 3x profits.
176. 5:38Here, we had three entry signals and
177. 5:40lost on all of them. Three?
178. 5:42>> A losing streak is rough.
179. 5:44>> As I mentioned at the start, the win
180. 5:46rate is 35 to 40%, so losing streaks are
181. 5:48part of this method.
182. 5:50But because the risk-reward ratio is
183. 5:52high, the expected value becomes
184. 5:54positive if you keep trading.
185. 5:56>> So it's okay even if the win rate is
186. 5:58under 40%. If the risk-reward ratio is 1
187. 6:00to 3, mathematically, excluding costs
188. 6:03like fees, a 25%win rate will let you
189. 6:05break even. You can win with a win rate
190. 6:08that low?
191. 6:08>> This is a case where I lost immediately
192. 6:10on the first entry, but then met the
193. 6:12conditions for a second entry and won by
194. 6:14going long.
195. 6:15>> Re-entered after losing and made it back
196. 6:17, I see.
197. 6:19Since my trading rule is that once
198. 6:21stopped out, the rationale is
199. 6:22invalidated, I usually don't do
200. 6:24consecutive entries like this, but
201. 6:26automating the strategy requires doing
202. 6:28these as well.
203. 6:29>> That upward-sloping graph is also from
204. 6:31when you took every entry based on the
205. 6:33conditions.
206. 6:34>> Exactly.
207. 6:35If I were to review all the entries and
208. 6:37tweak them, the win rate might improve
209. 6:39significantly.
210. 6:41But this time, I decided to release it
211. 6:42with almost no adjustments from me.
212. 6:45>> So, it's exactly as the AI created it?
213. 6:48I'm thinking of reviewing it again to
214. 6:50brush up on this strategy.
215. 6:53I'd love for everyone to give it a try.
216. 6:55I believe customizing it into your own
217. 6:57strategy makes the edge last longer, and
218. 6:59that experience can lead to discovering
219. 7:01new edges.
220. 7:02>> So instead of using what the AI made as
221. 7:04is, you make it your own weapon.
222. 7:06>> That's right.
223. 7:08I think turning it into your own weapon
224. 7:09is important for surviving in the market
225. 7:11.
226. 7:12Finally, let's look at the chart zoomed
227. 7:14out.
228. 7:15The marked spots are where the entry
229. 7:16signals appeared.
230. 7:18>> The signals are clustered in certain
231. 7:19areas.
232. 7:20>> Yes, they are concentrated during
233. 7:22uptrends, and barely appear during
234. 7:24downtrends or ranging periods.
235. 7:26Now, I'll share three weaknesses of this
236. 7:28strategy.
237. 7:29First, it only buys, so it only
238. 7:31functions during times when gold is in
239. 7:33an uptrend. Trades happen about 2 to 9
240. 7:36times a month, and there are periods
241. 7:38like from late 2026 to July where there
242. 7:40wasn't a single trade.
243. 7:41>> Are there times when no signals appear
244. 7:43for months? Second. Like that 3-loss
245. 7:45streak earlier, losses can pile up; in
246. 7:47the backtest, there were even 5 to 7
247. 7:48consecutive losses. Seven.
248. 7:50>> Losing streaks are hard on your
249. 7:51mentality. Third. In a strong uptrend,
250. 7:53you can win to some extent just by
251. 7:55buying without waiting for pullbacks, so
252. 7:57the extra edge of this method isn't
253. 7:59significant. The benefit of waiting for
254. 8:01pullbacks clearly appeared during
255. 8:02periods of weak growth. Also, after
256. 8:04finalizing the rules and double-checking
257. 8:06the charts from January to October 2026
258. 8:09, there were only 22 trades, and the
259. 8:11profits were skewed toward a few large
260. 8:13wins. So, with so few trades, it’s
261. 8:15possible it was just a coincidence.
262. 8:18>> We’ve looked at the actual charts, so
263. 8:20finally, I'll tell you what Reimu thinks
264. 8:21of this method.
265. 8:23>> I’ve been waiting for this.
266. 8:25>> Reimu hasn't used this method yet, but I
267. 8:27don't mean to say that it's a bad one.
268. 8:29In fact, I think it seems perfectly
269. 8:31usable.
270. 8:31>> So, are you going to start entering
271. 8:33trades when a signal appears starting
272. 8:35tomorrow?
273. 8:35>> I won't just enter without thinking
274. 8:37whenever a signal comes. I don't think
275. 8:38I'd use it for what you'd call a "
276. 8:40no-brainer" entry. There are two reasons
277. 8:42for that.
278. 8:43First, since many people can easily find
279. 8:45similar methods using AI, I think if
280. 8:47everyone uses the same approach, the
281. 8:49edge—or the advantage—of the method will
282. 8:52continue to disappear.
283. 8:55I previously mentioned that a bot used
284. 8:57by an acquaintance of mine would become
285. 8:59useless within a month at the soonest,
286. 9:01or a year at the latest, but if anyone
287. 9:02can create methods, that speed might
288. 9:04even accelerate.
289. 9:07So that's why you mentioned earlier that
290. 9:08it's important to make it your own
291. 9:10unique method.
292. 9:11Exactly.
293. 9:13Second, just because a graph trends
294. 9:15upward in a backtest doesn't guarantee
295. 9:17the same results in the future; the
296. 9:19graph at the beginning only shows past
297. 9:21performance, so it's actually possible
298. 9:22it could trend downward from next month
299. 9:24on.
300. 9:26>> I feel embarrassed for calling it the "
301. 9:28holy grail" at the beginning.
302. 9:30>> So, if Reimu were to use this method, I
303. 9:32would only enter when the signal aligns
304. 9:34with other factors, like higher
305. 9:36timeframe analysis, horizontal lines, or
306. 9:38chart patterns.
307. 9:40If you all decide to use it, I would
308. 9:42recommend that kind of approach.
309. 9:44>> Come to think of it, seasonally, it's
310. 9:46almost time for the Gold Anomaly. That's
311. 9:48right. Since this method pairs well with
312. 9:51uptrends, I think it's highly likely to
313. 9:52show its true potential during the Gold
314. 9:54Anomaly period.
315. 9:56It might be good to try aiming for it
316. 9:57actively.
317. 9:58>> Reimu, you should use this method for
318. 10:00anomalies too.
319. 10:01>> Reimu's favorite is still the usual
320. 10:03entry.
321. 10:04I think I’ll use this method as a
322. 10:06secondary backup.
323. 10:07And finally, the most important warning
324. 10:09I want to share.
325. 10:11I think scams involving automated
326. 10:12trading bots or tools showing
327. 10:14upward-sloping charts like this will
328. 10:16keep increasing.
329. 10:17So, I want you to be more careful than
330. 10:19ever.
331. 10:20>> Why do you think so?
332. 10:21>> We had the AI find this method entirely
333. 10:23on its own, didn't we?
334. 10:25The way you give instructions will
335. 10:27change the results, but those are just
336. 10:29minor differences.
337. 10:31Even though I'm a total amateur at
338. 10:33programming, I can easily build
339. 10:34automated trading tools just by asking
340. 10:36AI.
341. 10:38>> You can build them even without
342. 10:39programming skills?
343. 10:40>> Yes.
344. 10:41And it's an era where you can use AI to
345. 10:43create the websites to sell them and
346. 10:45even the promotional videos.
347. 10:48>> So you can whip up professional charts,
348. 10:50fancy sites, and convincing videos in no
349. 10:52time?
350. 10:53>> Exactly, so a pretty chart or a polished
351. 10:55website is no longer a reason to trust
352. 10:57something.
353. 10:59Anyone can create just the look of it.
354. 11:01And the tricky part is that it’s not
355. 11:03even a lie; it’s actually true.
356. 11:05>> That's true, there wasn't a single lie
357. 11:07in this method either.
358. 11:10>> Right, but even with this method, I
359. 11:11don't think for a second I'll earn money
360. 11:13forever; I even think it could turn
361. 11:15negative as early as next month.
362. 11:19To me, the credibility of automated
363. 11:21methods like this is low, but to
364. 11:22beginners who don't know much about FX
365. 11:24or those who are losing, it probably
366. 11:26looks like a Holy Grail.
367. 11:29They might even decide it's worth paying
368. 11:31100,000 or 200,000 yen for it.
369. 11:34>> At the start, Marisa, you were reacting
370. 11:36exactly like someone who would get
371. 11:37scammed.
372. 11:38>> I really want you to be careful of scams
373. 11:40so you don't get misled by this kind of
374. 11:41information.
375. 11:42>> I'll keep that in mind.
376. 11:44>> So, here is the summary for today. First
377. 11:46.
378. 11:47This method was created by AI, and I
379. 11:49haven't entered a single trade with it
380. 11:51yet.
381. 11:52Second, the rules are all determined by
382. 11:54numbers, so anyone can make the same
383. 11:56judgment. The validation graph trends
384. 11:58upward, so the method itself seems quite
385. 12:00usable. However, the win rate is 35 to
386. 12:0240%, and there will be losing streaks.
387. 12:04Third, since anyone can find similar
388. 12:06methods using AI, the edge will
389. 12:08disappear, and it could start trending
390. 12:10downward as early as next month. That’s
391. 12:12exactly why it’s important to customize
392. 12:14it into your own unique method. Fourth.
393. 12:16It’s a fact that it has been profitable
394. 12:18so far, so it all depends on how you use
395. 12:20it. If you use it, don’t just enter on
396. 12:21the signal; combine it with other
397. 12:23factors like market analysis. Gold
398. 12:25anomaly periods might be a good target,
399. 12:27but Reimu plans to use it strictly as a
400. 12:29secondary tool. Fifth, be extra cautious
401. 12:31of scams involving automated trading or
402. 12:33tools that show upward-trending graphs.
403. 12:35I’ve learned the method and understood
404. 12:37the risks of the AI era.
405. 12:39We’re in an age where you can even build
406. 12:41tools with AI, but it’s up to you to
407. 12:43decide how to use them and what to
408. 12:44believe.
409. 12:46Marisa will try validating it on her own
410. 12:48first, too.
411. 12:49>> I’ve put together a newly created
412. 12:50playlist in the description box.
413. 12:53Feel free to watch according to your own
414. 12:54weaknesses or what you want to improve.
415. 12:57That’s all for this time.
416. 12:59I hope you’ll relax and watch again next
417. 13:00time.
418. 13:04If you keep shining, the illusions
419. 13:05sleeping in your dreams will be in your
420. 13:07hands