I Tried to Break This Forex Edge. I Couldn't. — Transcript

Full transcript

1. 0:00So, one of the members in my paid
2. 0:02community claimed he found an edge in
3. 0:04dollar yen. Not an indicator, not a
4. 0:07course, a rule from an academic paper
5. 0:10that he trades live every fifth day for
6. 0:1235 minutes. This is his post.
7. 0:16By the way, if you want to be part of
8. 0:18this paid community, the link is below.
9. 0:21But, I also have my free YouTube
10. 0:23community, and they are, let's say,
11. 0:26intellectually different. And I want to
12. 0:28share a conversation with you. I'm doing
13. 0:31this to make you feel better.
14. 0:33Sometimes, you just feel like you're not
15. 0:35enough in this world. But, I promise
16. 0:37you, after reading this conversation,
17. 0:40you will.
18. 0:46>> [snorts]
19. 1:01[music]
20. 1:04[snorts]
21. 1:11>> Okay, back to the edge. I took his rules
22. 1:14and tested them on more than 10 years of
23. 1:16data against 10,000 random calendars,
24. 1:20and I had a second independent back test
25. 1:23rebuild it from scratch. It's one of the
26. 1:25very few strategies I couldn't break.
27. 1:28And believe me, I tried. Breaking other
28. 1:31people's strategies is pretty much my
29. 1:33whole personality at this point. My wife
30. 1:36says it's also the reason we don't get
31. 1:38invited to barbecues anymore.
32. 1:40First, what's a gotobi day?
33. 1:43In Japan, a lot of companies settle
34. 1:45their bills on days that end in five or
35. 1:48zero. The fifth, the 10th, the 15th, and
36. 1:51so on.
37. 1:52If that day is a weekend or a holiday,
38. 1:55it moves to the business day before. A
39. 1:57whole country paying its bills on fixed
40. 2:00days, on time.
41. 2:02Somewhere a CFO on his third marriage
42. 2:05just felt personally attacked. Japanese
43. 2:08importers pay their suppliers in
44. 2:09dollars. So, on those days they all need
45. 2:12dollars at the same time. And there's a
46. 2:14fixed moment when they get them. At 9:55
47. 2:18in the morning Tokyo time, the banks set
48. 2:20the so-called Tokyo fix, the rate their
49. 2:23customers pay for the whole day. Imagine
50. 2:26every company in Germany had to buy its
51. 2:28dollars at one price set at the same
52. 2:30minute on the same six days a month.
53. 2:33Somebody would notice.
54. 2:35Two Japanese economists did. Ito and
55. 2:38Yamada published a paper on it in 2017.
56. 2:42Dollar yen tends to rise into the fix on
57. 2:44these days and then give it back
58. 2:46afterwards. His rule is simple. On a
59. 2:49gotobi day, sell dollar yen at 9:55
60. 2:52Tokyo time. Stop loss 12 pips above the
61. 2:55entry. Close at 10:30. About 70 trades a
62. 2:59year, 35 minutes each, and you're flat
63. 3:02long before anyone in Europe is awake.
64. 3:0535 minutes is less time than the average
65. 3:07finance bro spends in the office
66. 3:09bathroom, and he isn't checking dollar
67. 3:11yen in there, either.
68. 3:13I tested it on dollar yen minute data
69. 3:15from January 2016 to October 2026.
70. 3:20I count the real spread at that time of
71. 3:22day and a bit of slippage on every
72. 3:24trade.
73. 3:25775
74. 3:27trades. On average, plus 2.35 pips per
75. 3:31trade after spread. The T statistic is
76. 3:345.5, which means this is very unlikely
77. 3:38to be luck. 58% of the trades win. For
78. 3:42comparison, most strategies I test on
79. 3:44this channel have a T statistic of
80. 3:46around zero and a Discord full of people
81. 3:49who swear they work.
82. 3:51Then I ran the exact same trade on every
83. 3:54other weekday. Same time, same stop,
84. 3:57same exit, just not on a gotobi day.
85. 4:00Result? Plus 0.01 pips. So, on a normal
86. 4:04day, after costs, this trade makes
87. 4:07nothing. The whole edge comes from the
88. 4:09calendar. So, if someone sells you a
89. 4:11secret time of day to trade, ask him
90. 4:14which day. Here's the average morning,
91. 4:17starting at 9:00 in Tokyo. On normal
92. 4:20days, dollar yen drifts up about one pip
93. 4:23into the fix, and then gives most of it
94. 4:25back. On Gotobi days, it's up almost two
95. 4:28pips at 9:55, and then it falls. By
96. 4:3210:30, it's 3.2 pips below the fix. It's
97. 4:36not a big move, but it shows up again
98. 4:39and again.
99. 4:40Every single year from 2016 to 2026 is
100. 4:44positive. The weakest was 2021 with plus
101. 4:48half a pip per trade. The best years
102. 4:50were 2022 and 2025 with more than four
103. 4:54pips.
104. 4:55Now, the usual objections. First, it's
105. 4:58been public since 2017, so it should be
106. 5:02arbitraged away by now. Since 2018, it's
107. 5:05plus 2.36 pips per trade, about the same
108. 5:09as in 2016 and 2017.
109. 5:12Apparently, the companies don't read the
110. 5:14Journal of International Money and
111. 5:16Finance. Neither do I, normally.
112. 5:19Second, maybe any set of days would look
113. 5:22this good. I drew 10,000 random
114. 5:25calendars with the same number of days.
115. 5:27Basically, I gave 10,000 monkeys a
116. 5:30calendar, and every single one of them
117. 5:32lost to a Japanese accounting
118. 5:34department. The best of them made 2.01
119. 5:37pips. Not one reached the real Gotobi
120. 5:40days. When I move the calendar one or
121. 5:42two days earlier, the edge is gone. One
122. 5:45day later, about 60% of it is still
123. 5:47there. So, it doesn't vanish the moment
124. 5:50the official day is over.
125. 5:52Third, my backtest could just be wrong.
126. 5:55It happens. You've caught me before, and
127. 5:57you'll catch me again. So, I gave only
128. 6:00the written rules to an independent
129. 6:02rebuild with its own code and its own
130. 6:05holiday calendar. It got 774 trades and
131. 6:09plus 2.39 pips. And our members own
132. 6:12numbers say two to three pips before
133. 6:14commission. So, three separate tests
134. 6:17land in the same place.
135. 6:19Fourth, and this one is fair. What does
136. 6:22it actually make? With a typical prop
137. 6:25firm commission of 0.7 pips, you keep
138. 6:28about 1.65 pips per trade. Risk half a
139. 6:32percent of the account per trade, like
140. 6:34he does, and that's roughly 5% a year
141. 6:37over 11 years. The worst drawdown was
142. 6:405.7%.
143. 6:42Two years, 2018 and 2021, ended slightly
144. 6:46negative after commission. If you came
145. 6:48here for 100% a month, you're on the
146. 6:51wrong channel, and the prop firms are
147. 6:53very happy to have you. He calls it a
148. 6:55reliable old small car, and I'd agree.
149. 6:58It's the most boring equity curve I've
150. 7:01ever shown on this channel, and I mean
151. 7:03that as a compliment. It's a real edge.
152. 7:06It's small, and it only works if you
153. 7:08actually take every trade at 9:55 Tokyo
154. 7:11time.
155. 7:12He also didn't test anything before
156. 7:14April 2020.
157. 7:16I did.
158. 7:17From 2016 to April 2020, it made plus
159. 7:211.84 pips per trade. So, it worked in
160. 7:24years he never looked at. That's about
161. 7:27the best evidence a backtest can give
162. 7:29you.
163. 7:30Here's what I like most about this one.
164. 7:32It didn't come from a guru or a course.
165. 7:35It came from someone in my community who
166. 7:38read a paper, built the rules carefully,
167. 7:41tested them properly, and then shared
168. 7:43it. That's exactly what the community is
169. 7:46for.
170. 7:47If you want to do the same, support my
171. 7:49work and get access to all code and
172. 7:51infrastructure and more than 50 hours of
173. 7:54educational video material, this
174. 7:56analysis included. You bring an idea, we
175. 7:59test it properly, and sometimes, like
176. 8:02today, it actually survives. The link is
177. 8:05in the description.
178. 8:07Thanks for watching, and I see you in
179. 8:09the next one.