Claude Opus 5 + MCP = New King of Algo Trading! — Transcript

Full transcript

1. 0:00The Anthropic team just released Claude
2. 0:02Opus 5, and this is huge, guys, because
3. 0:05this model is beating Fable 5 while
4. 0:08costing significantly less. In fact,
5. 0:10even on a $200 max subscription of
6. 0:13Anthropic, I was not really able to use
7. 0:15Fable 5 as much as I needed. Whenever I
8. 0:17ran a task for like 30 minutes to 1
9. 0:20hour, it would drain my usage very
10. 0:23quickly. But with Opus 5, I can get the
11. 0:25same quality for a significantly cheaper
12. 0:28price. And look at this benchmark. Like,
13. 0:30it's beating it in agenting terminal
14. 0:32coding, knowledge work, which is also
15. 0:34important for us because imagine if you
16. 0:36want to come up with some sort of
17. 0:38trading strategy idea, this model is
18. 0:40going to be able to help you better with
19. 0:41that. It's also scoring 30% for novel
20. 0:44problem solving, which is absolutely
21. 0:46crazy because even GPT-5.1.6 solved,
22. 0:49which was my favorite model until
23. 0:51yesterday, is a scoring 7.8%. So, this
24. 0:54is crazy, guys. So, in this video, as
25. 0:57always, I'm going to be testing the
26. 0:58model to see how good it is for writing
27. 1:01strategies, running back tests,
28. 1:03parameter optimizations, and Monte Carlo
29. 1:05simulations to ensure not only it finds
30. 1:07us good strategies, but also it ensures
31. 1:10they are not overfit. Before I move on,
32. 1:12I got to mention I am not a financial
33. 1:14advisor, and this video is for
34. 1:16educational purposes only. So, guys,
35. 1:18let's get started. As always, I'll be
36. 1:20using the Jesse framework for the algo
37. 1:22trading side of things using Python.
38. 1:24It's absolutely free to get started
39. 1:26with, and you can follow the
40. 1:27documentation to get it installed on
41. 1:29your machine. And if you prefer videos,
42. 1:31I also have those on my channel. So,
43. 1:33just go ahead and get started with that.
44. 1:35And assuming you've already done that,
45. 1:37all I need to do to get started is to
46. 1:39run the Jesse run command. There we go.
47. 1:42This is the address for the dashboard,
48. 1:43and this is the address for the MCP. So,
49. 1:45let's copy this. Now, in order to run
50. 1:48Claude code, we're going to have to run
51. 1:49it inside another terminal. And because
52. 1:51I'm on the Z editor, it allows us to do
53. 1:54it right from here. Now, this is another
54. 1:56terminal, very similar to this one, but
55. 1:59it is on the sidebar. So, I could just
56. 2:01run the Claude command in order to run
57. 2:03Claude code on my terminal right now,
58. 2:05but if you haven't added Jesse MCP into
59. 2:08Claude, you need to do it using this
60. 2:11command that is on the documentation of
61. 2:13Jesse. And you notice the address that I
62. 2:15copied was running on the port 9007. So,
63. 2:18instead of 9002, I have to enter that.
64. 2:20But for you, this is going to be the
65. 2:22default value. So, just go ahead, copy
66. 2:24this and run it here, and you should be
67. 2:26good to go. I've already done this, so
68. 2:28I'm not going to do that. I will just
69. 2:29instead run the Claude command. There we
70. 2:31go. Now, to ensure that I'm running on
71. 2:34the Opus 5, I'm going to run the /model
72. 2:37command, and here you can see that the
73. 2:39default is now using Opus 5. And the
74. 2:42effort is set to high, which I find it
75. 2:44to be more than enough for my use case.
76. 2:46All right, next we're going to have to
77. 2:47feed the prompt to the model. And in
78. 2:49case if you just want to copy and paste
79. 2:51it, I'm going to put it on GitHub and
80. 2:53link to the repository in the
81. 2:54description of the video. So, let's get
82. 2:56started. Hey there, I need you to do
83. 2:59research and develop a trend following
84. 3:01strategy for trading ETH USDT that
85. 3:04enters on pullbacks within an
86. 3:06established trend. Use supertrend as the
87. 3:08primary indicator, supported by an
88. 3:11EMA-based trend filter, and any other
89. 3:13confirmation indicator you see fit. Use
90. 3:16the 30-minute time frame as the main
91. 3:18trading time frame, and the 4 hours as
92. 3:20the anchor time frame. I need the
93. 3:22results of it to be good since the
94. 3:24beginning of the year until now. And
95. 3:26when I say good, I mean a sharp ratio of
96. 3:28at least 1.5. Risk 3% of the account per
97. 3:32each trade. Every time you develop a
98. 3:33strategy, validate the result using a
99. 3:36statistical significance test before
100. 3:38writing the full strategy to ensure the
101. 3:40entry rules are not pure noise. Only
102. 3:43proceed if the strategy's metrics
103. 3:45demonstrate genuine statistical
104. 3:46significance. Feel free to use
105. 3:48optimization to improve the results. At
106. 3:50the end, apply Monte Carlo simulations
107. 3:53to ensure the results are not overfit.
108. 3:55Continue until you find strategies that
109. 3:58fully meet the criteria requested. Do
110. 4:00not prompt me in the meanwhile. Good
111. 4:02luck. There we go. Let's hit enter. This
112. 4:05seems fine. And as you can see, it's
113. 4:07started reading the
114. 4:08agents.to.them.defile of the project,
115. 4:10which is going to tell it to use J S MCP
116. 4:13in order to write the strategy and do
117. 4:15the research. So, let's go and come back
118. 4:17a bit later. All right, so it did it
119. 4:18everything that I needed to do and it
120. 4:21performed it very quickly. So, I
121. 4:22basically just went for lunch, I came
122. 4:25back and it was ready. And as you can
123. 4:26see, it's also very verbose. And this is
124. 4:29helpful if you are a beginner. So, for
125. 4:31example, you can see exactly the path
126. 4:34that the AI is taking in order to
127. 4:36develop the strategy. And especially for
128. 4:39beginners, that is super helpful because
129. 4:41you want to learn. So, the first thing
130. 4:42that it did is that once it wrote the
131. 4:44strategy, it ran an RST test, which is
132. 4:48rule significance test, which it allows
133. 4:50us to know whether the entry rules of
134. 4:52the strategy were pure noise or luck or
135. 4:55was there an actual edge in it. So, it
136. 4:57does that before moving on with the rest
137. 5:00of the strategy such as the position
138. 5:02sizing, take profit and a stop loss.
139. 5:04Because if the entry rules of the
140. 5:05strategy don't have an actual edge, then
141. 5:08everything else that we do is pointless.
142. 5:09So, it did that to validate the results
143. 5:12that it found and then it continued
144. 5:14until it found a good strategy.
145. 5:16Here we can see it says version four is
146. 5:18profitable, which means it wrote four
147. 5:20versions until this point. There's
148. 5:22another V5. So, it tried out many
149. 5:25different variations of the same
150. 5:27strategy until at this point it was able
151. 5:29to find the ending result that we
152. 5:32wanted. So, a sharp ratio of 1.68, net
153. 5:35profit of 30%, a max drawdown of minus
154. 5:3717% and it executed 48 trades. So, this
155. 5:41looks good and it gave me URLs for the
156. 5:43final backtest, the RST test and the
157. 5:46Monte Carlo so we can check it out. But,
158. 5:48before I move on, I also asked the model
159. 5:50to do something else. I needed to add
160. 5:52charts to the strategy so that we can
161. 5:54visualize it better. And that's
162. 5:56literally what I told it. But, now that
163. 5:58I'm thinking, it would probably be
164. 5:59better if I added this to the prompt
165. 6:01that I gave the model in the first
166. 6:03place. So, let's just scroll down, and
167. 6:05here's the backtest results for what we
168. 6:08asked, and it also went ahead and tested
169. 6:10with other periods, such as this one,
170. 6:11which is for 2 and 1/2 year, and this
171. 6:13one is for 5 and 1/2 years. And of
172. 6:16course, these are super helpful if you
173. 6:18want to use the strategy to go live with
174. 6:20it. So, let's open this one first. Now,
175. 6:23one problem I have with it is that this
176. 6:25format that it is printing this, it
177. 6:27doesn't work. Okay, if I click on this,
178. 6:29we're going to get an error because this
179. 6:30ID is not complete. We also need this
180. 6:33one. So, let's ask it to fix this.
181. 6:34Please give me new URLs for everything,
182. 6:37not just the backtest, and also ensure
183. 6:39the URLs are on one line so I can just
184. 6:42click on it because the format that you
185. 6:44just printed them in a table doesn't
186. 6:46work. Now, while that's going, I want to
187. 6:48quickly remind you guys about our
188. 6:49Telegram. It's the fastest way to get
189. 6:51notified about my future work, whether
190. 6:53it's a new tutorial or a tool that I
191. 6:55create. Also, don't forget to check out
192. 6:56our free Discord, where more than 5,000
193. 6:59members like you and I are hanging out
194. 7:00there and helping out each other with
195. 7:02algo trading so we can all succeed
196. 7:04together. The links for both are down in
197. 7:05the description. All right, there we go.
198. 7:07So, let's open this one and also this
199. 7:09one and this. So, there we go. This is
200. 7:11the result of the backtest from the
201. 7:14beginning of the year until this month.
202. 7:16Now, if you see the dashboard being
203. 7:17different, that's because I'm running
204. 7:18version three of Jesse, which hasn't
205. 7:20been released yet, but it is coming in
206. 7:22the coming days. I have basically
207. 7:24redesigned everything and made
208. 7:25everything better. But anyways, here's
209. 7:27the equity chart, and this indigo equity
210. 7:30curve is the one for a strategy's
211. 7:32performance, and this one in orange is
212. 7:35the performance of ETH/USDT if you were
213. 7:37just buying and holding the asset. And
214. 7:40here's the chart for the worst five
215. 7:42drawdown periods, and as you can see,
216. 7:43it's not looking great. Like short, this
217. 7:46is awesome, and so is this one, but we
218. 7:48had a nasty drawdown period in here, and
219. 7:50this is the monthly return chart. It
220. 7:52executed 48 trades, the net profit is
221. 7:5530%, the max drawdown is -17, the win
222. 7:58rate is 31%, but the average win to loss
223. 8:01is really huge, so it is 3.8. The
224. 8:04average holding time is 38 hours, the
225. 8:07Sharpe ratio is 1.68, which is way above
226. 8:091.5 criteria that we set for the model.
227. 8:12And the average trades per month is
228. 8:14seven. So, overall, looking really good,
229. 8:16and the model was kind enough to also
230. 8:18execute it on other periods. So, this
231. 8:20one is for 2 and 1/2 year, and it
232. 8:22actually looks really good. So, here's
233. 8:25the equity curve chart, and as you can
234. 8:27see, it's going up nice and steady while
235. 8:29the actual underlying asset, which is
236. 8:31ETH, is in a range market. So, if you
237. 8:33were just holding ETH, you would not
238. 8:35have made a lot of money, but with this,
239. 8:37it would have been much better. The max
240. 8:39drawdown is still -17%, the win rate is
241. 8:4227%, but the average win to loss ratio
242. 8:45is 3.95.
243. 8:47So, guys, a strategy like this is very
244. 8:50difficult to trade mentally. So, from a
245. 8:53psychology side, most of us are not
246. 8:55going to be able to continue trading
247. 8:57something like this, because basically
248. 8:59every seven out of 10 trades that you
249. 9:01take are going to be losing ones. And
250. 9:02the Sharpe ratio is 1.35, which is still
251. 9:05good. And the average trades per month
252. 9:07is eight. So, overall, this is looking
253. 9:09really good. And the chart for the worst
254. 9:12five drawdown periods is also looking
255. 9:14better now. So, you see, guys, zooming
256. 9:16out is always the key, because in the
257. 9:19end, we're not going to be trading just
258. 9:20one strategy, right? So, even for this
259. 9:22period, for example, which is multiple
260. 9:24months, and the strategy wasn't doing
261. 9:26fine, if you can develop another
262. 9:28strategy which isn't super correlated to
263. 9:31this one, and if that one was doing fine
264. 9:33during this period, then we we have been
265. 9:35fine, because the other strategy was
266. 9:37making money while this one was in a
267. 9:39range. And when you execute multiple
268. 9:41strategies like this simultaneously,
269. 9:42you're going to get a better sharp ratio
270. 9:44and even a better max drawdown. So, that
271. 9:47is something to remember that do not
272. 9:49expect to just trade one strategy in the
273. 9:51end. Anyways, moving on, here's the
274. 9:53chart for the monthly returns of it and
275. 9:56as you can see, every year it is
276. 9:57profitable and in most months we are
277. 10:00also green, but not in all of them. And
278. 10:02we have months as bad as minus 11%. We
279. 10:05also have good ones as good as 17%.
280. 10:07Next, let's take a look at the results
281. 10:09for the 5 and 1/2 years of backtest and
282. 10:13as you can see, it still looks good even
283. 10:15though there are periods such as this
284. 10:17one where it was totally in a range. So,
285. 10:19the strategy's results wasn't super
286. 10:21awesome, but it wasn't that bad either.
287. 10:23So, it's not like it was falling a lot.
288. 10:25We also had a bad period here. So, even
289. 10:27though the price was going up, the
290. 10:28strategy wasn't performing well and we
291. 10:31were actually in a loss. So, this is
292. 10:32very important to remember that a
293. 10:35strategy like this could even lose money
294. 10:37in the short term even when the market
295. 10:39is going up and that is going to be very
296. 10:41difficult mentally. So, if I were
297. 10:42trading this strategy at this period, I
298. 10:45would have probably stopped and would
299. 10:46have lost on these profits over the next
300. 10:49few years. So, this is also really
301. 10:51important and another reason for why you
302. 10:53guys need a portfolio of strategies
303. 10:56instead of just trading one because it's
304. 10:58going to be a lot easier mentally if you
305. 11:00had other strategies running in this
306. 11:02period and you weren't allocating all of
307. 11:04your capital to just one single
308. 11:06strategy. Anyways, here's the drawdown
309. 11:08chart and this one was not only the
310. 11:11worst one, it was also the longest one.
311. 11:14And in fact, we can actually see how
312. 11:16long that was by taking a look at this.
313. 11:18So, max underwater period was 205 days.
314. 11:22So, close to a year. So, this was very
315. 11:24brutal if you were trading it during
316. 11:26that period. Anyways, so the max
317. 11:28drawdown is minus 20% which is a good
318. 11:30number. The total number of trades is
319. 11:32541.
320. 11:34The P&L would have been 607%.
321. 11:37The win rate was 28%. The sharp ratio
322. 11:40was 1.34. And the average trades per
323. 11:43month was eight. So, overall, I really
324. 11:45like this result. Even though so far I
325. 11:47just spent like an hour developing it
326. 11:49with this AI model. So, of course, I'm
327. 11:52going to keep working on it. But even as
328. 11:54it is, it does have a lot of potential.
329. 11:56Now, one more thing I want to show you
330. 11:57guys. Remember when I asked the model to
331. 11:59add charts to it. So, because I did
332. 12:01that, here I can click on the trade
333. 12:04chart button here and it's going to show
334. 12:06me an interactive chart which allows us
335. 12:09to see every single trade that it took.
336. 12:11So, for example, let's sort it based on
337. 12:13the best ones. And this one presumably
338. 12:16made 45%, right? So, let's click on it
339. 12:19and zoom in a little bit. So, we can see
340. 12:22exactly where it opened and where it
341. 12:23closed, right? But these indicator
342. 12:26values, these are the ones that we asked
343. 12:27it to add. So, here we can see the
344. 12:30regime, for example. So, when was it
345. 12:32minus one? So, these were ready for a
346. 12:34short position. And this was zero, so a
347. 12:37neutral territory as you can also see
348. 12:38the label here. But I want to minimize
349. 12:40this one to have more place for this.
350. 12:42And we also have the ADX, but I also
351. 12:44want to minimize this one. And here we
352. 12:47can see this blue line here is the
353. 12:494-hour EMA 200, which is the strategy
354. 12:51it's using. So, let's also toggle this.
355. 12:54And this orange one is the super trend
356. 12:56for the 4-hour time frame. So, let's
357. 12:58toggle this one also. And this red line
358. 13:01here is the stop loss. So, let's also
359. 13:04toggle this one. So, that is only going
360. 13:05to leave us with this purple line, which
361. 13:07is the super trend value on the
362. 13:1030-minutes time frame. So, let's zoom in
363. 13:12a little bit. Next, I'm going to go
364. 13:14click on here. So, now we can see it
365. 13:16went short here. And let's zoom out a
366. 13:19little bit. There we go. And if I hold
367. 13:21the mouse here again, we can see exactly
368. 13:24where where opened the trade and where
369. 13:26it closed it. So, let's also click on
370. 13:28this one, another one. Now, we can also
371. 13:31sort it based on the worst ones. So,
372. 13:33this one it lost 12%. Now, by the way,
373. 13:36this isn't 12% of your entire capital,
374. 13:38of course, because we were only risking
375. 13:392% of the capital per each trade. So,
376. 13:42this cannot be more than that, of
377. 13:43course, but this is just based on the
378. 13:46the price movement. So, I wouldn't
379. 13:48really look at this or care about it a
380. 13:50lot, but anyway, so let's just click
381. 13:52here. So, here we can see a losing
382. 13:54trade. So, let's zoom in. So, we went
383. 13:56short here, but the price actually went
384. 13:58up. Let's see another one and also
385. 14:00another one. All right, going back to
386. 14:02the terminal where we had the cloud
387. 14:04code, we can also see the results for
388. 14:06the significance test, for example. So,
389. 14:08let's give this one a
390. 14:10try. There we go. So, here's the chart.
391. 14:12Because the P value is lower than this
392. 14:14value here, we're going to say the entry
393. 14:16rules of the strategy has statistical
394. 14:18significance. So, I am a little bit over
395. 14:20simplifying this because explaining how
396. 14:23this algorithm works requires its own
397. 14:25video, but basically, this is all you
398. 14:28need. So, if the results is saying that
399. 14:30it is significantly significant, that's
400. 14:32it. Just move on to the other parts of
401. 14:34the strategy. And last, but definitely
402. 14:36not least, the Monte Carlo simulation is
403. 14:38the most important piece of the puzzle
404. 14:40because sure, we got these results and
405. 14:43they were good, but you want to ensure
406. 14:44that it isn't overfit. And as you can
407. 14:46see in the chart, the orange line is the
408. 14:49original backtest equity curve, and we
409. 14:51can see that it is ending in right in
410. 14:53the middle. And the simulations are way
411. 14:55above it. So, this is a really good sign
412. 14:57because if this one, for example, was
413. 14:59just here, like way at the top, it would
414. 15:02have been a clear sign that we were just
415. 15:04being lucky during those backtests,
416. 15:07which is not a good sign if you want to
417. 15:08take the strategy live. And if you also
418. 15:11read this table here, we can see the
419. 15:12Sharpe ratio of the original was 1.62,
420. 15:15but the median number is 1.93, and the
421. 15:18best 5% is 4.36. So, usually, I want the
422. 15:22original back test sharp ratio to be
423. 15:25near median. So, something such as two
424. 15:27or even 2.4 maybe, that would have been
425. 15:30fine for me. So, I would have said that
426. 15:31that is not overfit. But, as you can
427. 15:33see, this case is even less than the
428. 15:35median number. So, this is a very good
429. 15:38sign and maybe, just maybe, this is also
430. 15:41the same reason why it's performing so
431. 15:42well during the bigger periods. Now, I'm
432. 15:45going to move on, but if you guys want
433. 15:47to take this strategy live, I also
434. 15:49suggest try Monte Carlo simulations on
435. 15:51these bigger periods. Now, we can also
436. 15:53see that the agent wrote us a complete
437. 15:56report file in the markdown format. So,
438. 15:58we can see what was the strategy's name
439. 16:01and what was the parameters that it
440. 16:02used, what was the target that it
441. 16:04achieved, and how it did it. It also
442. 16:06gives us a pretty good summary of
443. 16:08everything that just went on. Now, let's
444. 16:10also take a look at the strategy's code.
445. 16:12So, we can see a really huge section of
446. 16:16the explanation. We can just skip all of
447. 16:18it. Here's the percentage that it risk
448. 16:20per each trade. It defined the anchor
449. 16:23time frame, which is always a good idea
450. 16:25for determining the bigger trend of the
451. 16:27market and the ATR indicator. It also
452. 16:29wrote some useful comments such as this
453. 16:31one, data and signal. We can see the
454. 16:33anchor regime, which I'm guessing is the
455. 16:36bigger trend, and it is using the super
456. 16:38trend, and it is passing the anchored
457. 16:40candle. So, pretty standard stuff. It is
458. 16:43also using the before method of Jesse in
459. 16:46order to store some variables to use
460. 16:49later. Now, I don't think this is
461. 16:51exactly the best syntax. I think it
462. 16:53wrote things a little bit more
463. 16:54complicated than what I personally would
464. 16:56have, but it is fine. So, moving on, it
465. 16:59also defined another method called the
466. 17:01entry signal, which is responsible to
467. 17:04tell us whether we want to open a long
468. 17:06or a short position. And then, inside
469. 17:08the should long and should short methods
470. 17:10of Jesse, it is simply returning the
471. 17:13value of this signal. So, it's saying
472. 17:15that if it's one, we want to go long. If
473. 17:17it's minus one, we want to go short.
474. 17:19However, this is actually a pretty good
475. 17:21syntax because it is just making the
476. 17:23decision if you want to open a long or
477. 17:25short position, but it's not actually
478. 17:27executing the orders. Next, inside the
479. 17:30go long and go short methods of Jesse,
480. 17:32we can see it's defining the entry
481. 17:34order, which is being the current price,
482. 17:36the stop loss order, and it is also
483. 17:38storing the stop loss order so that we
484. 17:41can use it later, and it is submitting
485. 17:43the buy and the sell orders with this
486. 17:45syntax. Now, because the results are
487. 17:47good, I'm going to give it a pass, but
488. 17:49in the actual prompt that we gave the
489. 17:51agent, I did not want it to open the
490. 17:54positions using market orders, and that
491. 17:56is why I said that I want to go long
492. 17:58during a pullback. So, what I meant by a
493. 18:00pullback was entering via a limit order.
494. 18:03Now, maybe I should have been more
495. 18:05specific with it, but this isn't exactly
496. 18:07what I had in mind. But anyways, moving
497. 18:09on. So, once it opens the position, it
498. 18:12is submitting the stop loss order, and
499. 18:14that is was storing the sub price here.
500. 18:17So, this is also good. And notice that
501. 18:20it did not define a take profit order,
502. 18:23and instead it's using the update
503. 18:24position method of Jesse, which is
504. 18:26executed after every single candle
505. 18:29closes to keep updating the stop loss
506. 18:32order, and also saying that if the
507. 18:35regime is no longer with us, we want to
508. 18:37liquidate the current position. So,
509. 18:39basically, this is using a trailing a
510. 18:41stop kind of take profit, which is
511. 18:43pretty a standard if you want to take
512. 18:45the most out of a trend. And I'm
513. 18:48guessing this is why the average win to
514. 18:50loss ratio of the strategy was so high
515. 18:52while the win rate was low. And finally,
516. 18:55we have the update chart method of
517. 18:57Jesse, which is actually new inside the
518. 19:00version three, which isn't released yet.
519. 19:01Again, it's coming in the coming days.
520. 19:03What it does is that it adds some
521. 19:05indicators to the chart so we can view
522. 19:07it inside the interactive chart. Now, we
523. 19:10actually had this feature before, but
524. 19:12this one is just the upgraded version of
525. 19:14that and it also works during live
526. 19:17trading. So, not just the back test,
527. 19:18which is the limitation that we had
528. 19:19before. But anyways, moving on, we also
529. 19:21have the upper parameters, which is
530. 19:23defined, such as the stop ATR, trailing
531. 19:26ATR, the ADX mean, which is for the
532. 19:28threshold of the ADX and things like
533. 19:30that. So, this is the syntax that we use
534. 19:32inside the JSE framework in order to run
535. 19:35optimization to find the best
536. 19:37parameters. As always, I did submit this
537. 19:39strategy on our strategies index page on
538. 19:41our website. So, if you want to check it
539. 19:43out on other trading periods and see the
540. 19:45metrics for those, you can check out
541. 19:47this page and I'm going to link to it in
542. 19:49the description of the video. You can
543. 19:50also check out this page for other
544. 19:52strategies submitted either by me or
545. 19:54other members of our community. I forgot
546. 19:57to mention you can also download the
547. 19:58source code of the strategy by clicking
548. 20:00here. Anyways, if you enjoyed the video,
549. 20:02please make sure to give it a like and
550. 20:04post a comment to let me know what you
551. 20:05think. It helps me out a lot and don't
552. 20:08forget to subscribe if you haven't
553. 20:09already because I will be creating more
554. 20:11videos just like this one in the future.
555. 20:13Thank you so much for watching. I'll see
556. 20:15you in the next one.
557. 20:18>> [music]