Skip to content
#small language model Open access

Round Numbers Do Not Attract Price: Levels Merely Shifted by 25 Pips Are Touched More Often in All Eight Cells, and the Close Sits 0.2186 Pips Farther from a Round Number Than Chance Would Put It [F011]

Sep 2026 · Zenodo (CERN European Organization for Nuclear Research)

Abstract

「キリのいい数字は意識される」「大台には注文が溜まる」という言い方がある。だが「キリ番に触れた」という観察だけでは何も言えない。キリ番は 50 pips ごとに必ずあるので、必ずどれかが近くにあり、近ければ触れる。決着させるには、同じ間隔で意味の無い水準を置けばよい。キリ番を 25 pips ずらすと、間隔は 50 pips のまま、キリのよさだけが消える。 四銘柄(USDJPY・EURUSD・GBPUSD・XAUUSD)の2019年から2024年までの1時間足で、その日の始値から上下それぞれに最も近い水準を取り、その日のうちに触れた率を測った。キリ番の触れた率の中央は 68.1877%、25 pips ずらした偽の水準は 70.3728% で、差は −2.1851 ポイント、偽の方が高い。しかも八つのセル(四銘柄 × 上下)すべてで偽の方が高く、偽の水準の方が始値から遠い(距離の中央 0.2142 対 0.2085)にもかかわらずである。触れたあと反対へ真の値幅の半分だけ戻る率も、キリ番 34.1199%・偽 33.9213% で差は +0.1986 ポイントであり、「キリ番では跳ね返る」も成立しない。 距離の五分位に切って同じ帯の中で比べると、キリ番の方が高かったのは 20 帯のうち 6 で、偶然なら 10.0 なので偶然を下回る。差の中央は −1.7972 ポイントだが、幅は −7.8066 から +5.0979 と 13 ポイント近くあり、符号も帯ごとに入れ替わる。これは効果ではなく、一帯あたり 300 件前後という件数の揺らぎである。 終値と最も近い水準の距離も測った。刻みが 50 pips なので、終値が一様に散らばるなら平均は 12.5 pips になる。実測はキリ番について 0.2186 pips 遠く、偽の水準については 0.2186 pips 近い。引き寄せがあるならキリ番との距離だけが 12.5 より小さくなるはずで、符号が逆に出ている。ただし 0.2186 pips はスプレッド(0.30 から 3.67 pips)より小さいので、本稿はこれを「キリ番が価格を遠ざける」とは読まない。読むべきは「引き寄せが見つからなかった」であって「反発が見つかった」ではない。 水準に触れた足に信号を置いて先の値動きを測ると、方向の情報の中央はキリ番 −0.6688 pips・偽 +0.7658 pips で、日次クラスタの t の絶対値が 2 以上のセルはキリ番 4 のうち 3・偽 4 のうち 4 である。偽の方が「有意」になっているが、これは偽に情報があるという意味ではなく、両方とも「水準に触れたあと跳ね返りに賭けると負ける」という同じ形を出しているにすぎない。足の中まで独立な正規増分で作った合成系列で同じ測定をすると、触れた率の差の中央は −0.5291 ポイントで、合成乱数には注文も心理もキリ番も無いのに同じ率が出る。しかも乱数でもキリ番と偽の差が −0.4692 から +1.2982 と散らばり、その散らばりは実測の差 −2.1851 と同じ桁である。時間帯を重ならない五区分に分けても、有意なセルは 15 のうち 1(偶然ならおよそ 0.7)である。 この論文の位置づけ:新しい数学定理も新しい法則も主張しない。価格の刻み、真の値幅、ランダムウォークはすべて既知である。特定の教材やその著者、販売元は主題ではなく、一切名指していない。売買の成績(手数料・約定・資金管理を含む)は測っていない。刻みを 50 pips と 100 pips に固定しており、10 pips 刻みや、銘柄ごとに意識されるとされる特定の水準は扱っていない。そして注文板を見ていない。「大台に注文が溜まる」という主張のうち、板に実際に注文があるかどうかは価格の足からは測れない。本稿が否定したのは価格の側に現れる引き寄せであって、板の中身ではない。測定は日中・少数銘柄・四銘柄六年に限られ、月次の時系列モメンタムのようにより長い保有で文書化されている結果を否定するものではない。 作成にあたって:本稿の着想と内容は、著者自身の考察に基づくものです。文章の構成整理や英訳、集計に用いたスクリプトの作成には AI(大規模言語モデル)の助力を得ました。最終的な内容の解釈や誤りがあれば、それらはすべて著者の責に帰します。お気づきの点があれば、ご教示いただければ幸いです。 ----- There is a way of speaking that says "round figures are watched" and "orders pile up at the big figure". But the observation "price touched a round number" says nothing on its own: a round number exists every 50 pips, so one is always near, and what is near gets touched. The question settles by placing meaningless levels on the same grid. Shifting the round numbers by 25 pips keeps the spacing at 50 pips and removes only the roundness. Using one-hour bars for four series (USDJPY, EURUSD, GBPUSD, XAUUSD) from 2019 to 2024, the nearest level above and below each day's open was taken and the share of days on which it was touched was measured. The median touch rate is 68.1877% for round numbers and 70.3728% for the fake levels shifted by 25 pips, a difference of −2.1851 points with the fake higher. The fake was higher in all eight cells (four series times two sides), and this although the fake levels sit farther from the open (median distance 0.2142 against 0.2085). The share retracing half a true range after a touch is 34.1199% for round numbers and 33.9213% for the fakes, a difference of +0.1986 points, so "price bounces at round numbers" does not hold either. Cutting into quintiles of distance and comparing within each band, the round numbers are higher in only 6 of 20 bands against 10.0 by chance. The median difference is −1.7972 points, but the range runs from −7.8066 to +5.0979 — nearly 13 points — and the sign changes from band to band. That is not an effect but the scatter of about 300 observations per band. The distance from the close to the nearest level was measured too. With a 50 pip grid, a uniformly scattered close sits a mean of 12.5 pips away. The measurement is 0.2186 pips farther for round numbers and 0.2186 pips nearer for the fakes. Attraction would make only the distance to round numbers smaller than 12.5; the sign comes out the other way. But 0.2186 pips is smaller than the spread (0.30 to 3.67 pips), so this paper does not read it as "round numbers repel price". What it reads is that no attraction was found, not that repulsion was found. Placing a signal on the bar that touches a level, the median directional information is −0.6688 pips for round numbers and +0.7658 pips for the fakes, and the daily-clustered t reaches 2 in absolute value in 3 of 4 cells for round numbers and 4 of 4 for the fakes. The fakes coming out "significant" does not mean they carry information: Dd is strongly negative for both, and both produce the same shape — betting on the bounce after a touch loses. On synthetic series built from independent normal increments down to the movement inside each bar, the median difference in touch rate is −0.5291 points: a random walk has no orders, no psychology and no round numbers, and gives the same rates anyway. Moreover, even on the random walk the gap between round and fake scatters from −0.4692 to +1.2982, of the same order as the measured −2.1851. Splitting into five non-overlapping sessions gives 1 significant cell of 15, against about 0.7 by chance. Scope of this paper: no new mathematical theorem and no new law is claimed. Price grids, the true range and random walks are all known. No real course, author, or vendor is the subject here, and none is named. Trading performance, including commissions, fills and money management, is not measured. The grid is fixed at 50 and 100 pips, so a 10 pip grid and particular levels said to matter in a particular instrument are not treated. And the order book is not observed: whether orders rest at a big figure cannot be measured from price bars. What this paper denies is attraction that shows up on the price side, not the contents of the book. The measurement is intraday and covers four instruments over six years, and does not contradict results documented at longer holding periods, such as monthly time-series momentum. Acknowledgement: The ideas and content of this paper stem from the author's own considerations. Assistance from an AI (a large language model) was used for structuring, English translation, and writing the scripts used for the measurement. Any remaining errors or misinterpretations are solely the author's. Feedback and corrections are sincerely appreciated.

View source

Similar papers

#small language model Dataset Open access Oct 2026

Socratic guiding questions in synthetic arithmetic data: matched LoRA runs (revision v2)

Supporting data, adapters, predictions and code for the article *Low-Cost LoRA Fine-Tuning of Small Language Models for Multi-Step Arithmetic Reasoning* by Jake O'Grady, Asena Isik Gürhan, Chee Fong Ting and Effirul Ramlan (University of Galway). We generated 20,000 GSM8K-derived arithmetic problems with step-by-step s...

O'Grady, Jake, Gürhan, Asena Isik, Chee, Fong Ting et al. · 465 citations
#computer vision Open access Jun 2016

Software Development in Startup Companies: The Greenfield Startup Model

The results are packaged in the Greenfield Startup Model (GSM), which explains the priority of startups to release the product as quickly as possible, and the need to shorten time-to-market, by speeding up the development through low-precision engineering activities.

Carmine Giardino, Nicolò Paternoster, M. Unterkalmsteiner et al. · 178 citations · ⚡14
#computer vision Open access Oct 2016

Software Startups - A Research Agenda

Software startup companies develop innovative, software-intensive products within limited timeframes and with few resources, searching for sustainable and scalable business models.

M. Unterkalmsteiner, P. Abrahamsson, Xiaofeng Wang et al. · 157 citations · ⚡17
#machine learning Review Open access Oct 2016

“Failures” to be celebrated: an analysis of major pivots of software startups

This study conducts a case survey study based on the secondary data of the major pivots happened in 49 software startups, and demonstrates that customer need pivot is the most common among all pivot types.

Sohaib Shahid Bajwa, Xiaofeng Wang, Anh Nguyen-Duc et al. · 127 citations · ⚡15
#computer vision Review Open access May 2015

A survey study on major technical barriers affecting the decision to adopt cloud services

The comparison of adopter and non-adopter sample reveals three potential adoption inhibitor, security, data privacy, and portability, which underlines the importance of the technical and security perspectives for research investigating the adoption of technology.

Nattakarn Phaphoom, Xiaofeng Wang, S. Samuel et al. · 111 citations · ⚡8
#computer vision Open access Feb 2018

Lean Internal Startups for Software Product Innovation in Large Companies: Enablers and Inhibitors

This study investigates how Lean internal startup facilitates software product innovation in large companies and identifies its enablers and inhibitors, and shows the potential of the method-in-action framework to investigate the Lean startup approach in non-startup context.

Henry Edison, Nina M. Smørsgård, Xiaofeng Wang et al. · 78 citations · ⚡6

Related blog posts

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.