📐

斐波那契（Fibonacci）数列在外汇交易中的应用（第一部分）：探究价格与时间的关系·综合运用

（3/3）·从10万次波动的统计验证到落地下单，前两步铺垫的数列关系终于能直接用在盘面

含代码示例实战向第 3/3 篇

很多人把斐波那契回撤画完就等价格自己回来，却从没验证过自己品种的时间窗口是否也吃这套数列。前两篇拆了原理和统计，这一篇直接把验证过的模式接进你的交易流程。

◍ 把比率识别变成可下单的预测

光在图表上认出斐波那契比率还不够，真正值钱的是把它转成带置信度的下一步预判。我们做了一套算法，不只数某个比率出现的频次，还把精确度、耗时和当时市场状态一起喂进去，每次出信号都附带一个历史回测得到的置信水平。回测里有个硬数据：系统标出「比率成形概率偏高」的情境，约 72% 后续价格波动真的摸到了预测位。更有意思的是，当价格维度和时间维度同时撞上 0.618 这种关键比率，准头会跳到 85%——比如观测到价格在 2 小时内走了 0.00273 点后，双维共振下的命中率就到了这个数。外汇和贵金属是高杠杆品种，这种统计优势不等于稳赢，实盘仍可能因流动性断裂而失效。跨品种校验也做了：欧美小时图之外，其他热门货币对同样跑出类似关系，只是准确度随波动率和主趋势强弱浮动。底下这段是预测函数的核心骨架，注意它用 confidence_threshold 卡掉低精度模式，再按比率分组算目标位。

MQL5 / C++

<span class="keyword">def</span> predict_next_movement(movements, patterns, time_patterns, confidence_threshold=<span class="number">class="num">0.95</span>):
&nbsp;&nbsp;&nbsp;&nbsp;predictions = []
&nbsp;&nbsp;&nbsp;&nbsp;last_movement = movements[-<span class="number">class="num">1</span>]
&nbsp;&nbsp;&nbsp;&nbsp;last_price = last_movement[<span class="class="type">class="kw">string">&class="macro">#x27;end_price&class="macro">#x27;</span>]
&nbsp;&nbsp;&nbsp;&nbsp;last_movement_size = last_movement[<span class="class="type">class="kw">string">&class="macro">#x27;movement&class="macro">#x27;</span>]
&nbsp;&nbsp;&nbsp;&nbsp;
&nbsp;&nbsp;&nbsp;&nbsp;<span class="comment"># High-precision pattern analysis</span>
&nbsp;&nbsp;&nbsp;&nbsp;high_accuracy_patterns = [p <span class="keyword">for</span> p <span class="keyword">in</span> patterns
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">if</span> p[<span class="class="type">class="kw">string">&class="macro">#x27;type&class="macro">#x27;</span>] == <span class="class="type">class="kw">string">&class="macro">#x27;price_ratio&class="macro">#x27;</span>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">and</span> (<span class="number">class="num">1</span> - p[<span class="class="type">class="kw">string">&class="macro">#x27;accuracy&class="macro">#x27;</span>]) &gt;= confidence_threshold]
&nbsp;&nbsp;&nbsp;&nbsp;
&nbsp;&nbsp;&nbsp;&nbsp;<span class="comment"># Group patterns by ratios</span>
&nbsp;&nbsp;&nbsp;&nbsp;ratio_groups = {}
&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">for</span> pattern <span class="keyword">in</span> high_accuracy_patterns:
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ratio = pattern[<span class="class="type">class="kw">string">&class="macro">#x27;ratio_name&class="macro">#x27;</span>]
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">if</span> ratio <span class="keyword">not</span> <span class="keyword">in</span> ratio_groups:
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ratio_groups[ratio] = []
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ratio_groups[ratio].append(pattern)
<span class="keyword">for</span> ratio_name, ratio_value <span class="keyword">in</span> fib_ratios.items():
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;patterns_with_ratio = ratio_groups.get(ratio_name, [])
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">if</span> <span class="keyword">not</span> patterns_with_ratio:
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">class="kw">continue</span>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="comment"># Analyze movement direction</span>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;up_count = <span class="built_in">sum</span>(<span class="number">class="num">1</span> <span class="keyword">for</span> p <span class="keyword">in</span> patterns_with_ratio
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">if</span> p[<span class="class="type">class="kw">string">&class="macro">#x27;movement2&class="macro">#x27;</span>] &gt; p[<span class="class="type">class="kw">string">&class="macro">#x27;movement1&class="macro">#x27;</span>])
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;down_count = <span class="built_in">len</span>(patterns_with_ratio) - up_count
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="comment"># Calculate probable target levels</span>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;target_levels = []
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">for</span> pattern <span class="keyword">in</span> patterns_with_ratio:
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">if</span> pattern[<span class="class="type">class="kw">string">&class="macro">#x27;movement1&class="macro">#x27;</span>] &gt; <span class="number">class="num">0</span>:
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;level = last_movement_size * pattern[<span class="class="type">class="kw">string">&class="macro">#x27;movement2&class="macro">#x27;</span>] / pattern[<span class="class="type">class="kw">string">&class="macro">#x27;movement1&class="macro">#x27;</span>]
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;target_levels.append(level)
<span class="comment"># Adjust forecasts based on time patterns</span>
&nbsp;&nbsp;&nbsp;&nbsp;time_patterns_high_accuracy = [p <span class="keyword">for</span> p <span class="keyword">in</span> time_patterns
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; <span class="keyword">if</span> (<span class="number">class="num">1</span> - p[<span class="class="type">class="kw">string">&class="macro">#x27;accuracy&class="macro">#x27;</span>]) &gt;= confidence_threshold]
&nbsp;&nbsp;&nbsp;&nbsp;
&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">for</span> pred <span class="keyword">in</span> predictions:
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;matching_time_patterns = [p <span class="keyword">for</span> p <span class="keyword">in</span> time_patterns_high_accuracy
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">if</span> p[<span class="class="type">class="kw">string">&class="macro">#x27;ratio_name&class="macro">#x27;</span>] == pred[<span class="class="type">class="kw">string">&class="macro">#x27;ratio&class="macro">#x27;</span>]]
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<span class="keyword">if</span> matching_time_patterns:

用历史时段吻合度给预测加权

在价格行为模型里，同一形态在不同交易时段的表现差异很大。把当前候选形态与历史同时段出现的相似模式做匹配后，可以取这些历史模式准确率的均值，用来动态修正当下预测的置信度。上面这段逻辑先算 matching_time_patterns 里每条历史模式准确率偏差（1 - accuracy）的均值 avg_time_accuracy，再用 (1 + avg_time_accuracy)/2 作为乘数放大或削弱 pred['confidence']。若历史吻合度普遍高，置信度会向 1 靠拢；偏差大则被压低。同时用 np.mean 取这些历史模式的 duration2 均值，写回 pred['expected_duration']，给持仓时间一个基于同时段统计的参考值。外汇与贵金属波动受时段流动性影响明显，这类加权在高风险品种上仅作概率参考，开 MT5 把这段接进你的模式匹配函数即可验证。

MQL5 / C++

    avg_time_accuracy = np.mean([class="num">1 - p[&class="macro">#x27;accuracy&class="macro">#x27;] for p in matching_time_patterns])
    pred[&class="macro">#x27;confidence&class="macro">#x27;] *= (class="num">1 + avg_time_accuracy) / class="num">2
    pred[&class="macro">#x27;expected_duration&class="macro">#x27;] = np.mean([p[&class="macro">#x27;duration2&class="macro">#x27;] for p in matching_time_patterns])

「把这条线请下神坛」

斐波那契叠加机器学习跑出的“时间共振”，本质是在价格空间与时间的双重轴上同时撞上黄金比率的同步摆动。作者在回测框架里给出的信号识别函数，在特定品种与时段组合下预测精度明显抬升，但这仍只是概率优势，不是市场律令。外汇与贵金属杠杆高、跳空频繁，把这类算法直接当进场圣旨极易被噪声反噬。更务实的做法是：用 MT5 接 Python 把附带的 FiboPattern_4 逻辑复刻成指标，只观察共振窗口是否重复出现，再决定仓位。数学和谐听起来像宇宙之舞，落到盘面上不过是一串可验证的 0.618 与 1.618 重合点。把它当显微镜，而非水晶球，才不会在下次背离时怀疑人生。

让小布替你跑这套验证

小布盯盘的 AIGC 已内置斐波那契价格-时间联合扫描，打开对应品种页就能看到近期波段是否落在数列窗口，不必自己写脚本抓10万根 K 线。

←

斐波那契（Fibonacci）数列在外汇交易中的应用（第一部分）：探究价格与时间的关系·进阶篇

常见问题

统计上时间窗口先到位、价格再碰回撤水平的组合，信号质量高于单一维度；可先框时间再等价格，避免提前进场。

能。小布在品种页输出数列共振提示，并标注历史命中概率，你只需决定是否跟，重复劳动交给它。

不一样。外汇与贵金属虽都受数列影响，但波幅结构不同，时间步长需分别拟合，切勿套用同一套参数。

以相邻数列时间单元的外侧极值作参考边界，结合 ATR 倍数，倾向宽于普通支撑阻力止损，外汇贵金属高风险需留缓冲。

斐波那契（Fibonacci）数列在外汇交易中的应用（第一部分）：探究价格与时间的关系·综合运用

◍ 把比率识别变成可下单的预测

用历史时段吻合度给预测加权

「把这条线请下神坛」

常见问题

相关阅读