nautilus_analysis/statistic.rs
1// -------------------------------------------------------------------------------------------------
2// Copyright (C) 2015-2026 Nautech Systems Pty Ltd. All rights reserved.
3// https://nautechsystems.io
4//
5// Licensed under the GNU Lesser General Public License Version 3.0 (the "License");
6// You may not use this file except in compliance with the License.
7// You may obtain a copy of the License at https://www.gnu.org/licenses/lgpl-3.0.en.html
8//
9// Unless required by applicable law or agreed to in writing, software
10// distributed under the License is distributed on an "AS IS" BASIS,
11// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12// See the License for the specific language governing permissions and
13// limitations under the License.
14// -------------------------------------------------------------------------------------------------
15
16use std::{collections::BTreeMap, fmt::Debug};
17
18use nautilus_model::{orders::Order, position::Position};
19
20use crate::Returns;
21
22const IMPL_ERR: &str = "is not implemented for";
23
24/// Trait for portfolio performance statistics that can be calculated from different data sources.
25///
26/// This trait provides a flexible framework for implementing various financial performance
27/// metrics that can operate on returns, realized PnLs, orders, or positions data.
28/// Each statistic implementation should override the relevant calculation methods.
29#[allow(unused_variables)]
30pub trait PortfolioStatistic: Debug {
31 type Item;
32
33 /// Returns the name of this statistic for display and identification purposes.
34 fn name(&self) -> String;
35
36 /// Calculates the statistic from time-indexed returns data.
37 ///
38 /// # Panics
39 ///
40 /// Panics if this method is not implemented for the specific statistic.
41 fn calculate_from_returns(&self, returns: &Returns) -> Option<Self::Item> {
42 panic!("`calculate_from_returns` {IMPL_ERR} `{}`", self.name());
43 }
44
45 /// Calculates the statistic from realized profit and loss values.
46 ///
47 /// # Panics
48 ///
49 /// Panics if this method is not implemented for the specific statistic.
50 fn calculate_from_realized_pnls(&self, realized_pnls: &[f64]) -> Option<Self::Item> {
51 panic!(
52 "`calculate_from_realized_pnls` {IMPL_ERR} `{}`",
53 self.name()
54 );
55 }
56
57 /// Calculates the statistic from order data.
58 ///
59 /// # Panics
60 ///
61 /// Panics if this method is not implemented for the specific statistic.
62 #[allow(dead_code)]
63 fn calculate_from_orders(&self, orders: Vec<Box<dyn Order>>) -> Option<Self::Item> {
64 panic!("`calculate_from_orders` {IMPL_ERR} `{}`", self.name());
65 }
66
67 /// Calculates the statistic from position data.
68 ///
69 /// # Panics
70 ///
71 /// Panics if this method is not implemented for the specific statistic.
72 fn calculate_from_positions(&self, positions: &[Position]) -> Option<Self::Item> {
73 panic!("`calculate_from_positions` {IMPL_ERR} `{}`", self.name());
74 }
75
76 /// Calculates the statistic from time-indexed strategy returns relative to a benchmark.
77 ///
78 /// Defaults to `None`; only benchmark-relative statistics (beta, alpha, information
79 /// ratio, tracking error, Treynor ratio) override this method. The `None` default
80 /// lets analyzer loops filter results by `Option` — non-benchmark statistics are
81 /// simply skipped, as `get_performance_stats_general` already does with
82 /// `calculate_from_positions` results — rather than panicking.
83 fn calculate_from_returns_with_benchmark(
84 &self,
85 returns: &Returns,
86 benchmark: &Returns,
87 ) -> Option<Self::Item> {
88 None
89 }
90
91 /// Aligns two returns series onto a common daily grid.
92 ///
93 /// Both series are first downsampled to daily bins (geometric compounding within each
94 /// UTC day), then inner-joined on shared timestamps. Timestamps present in only one
95 /// series are dropped (not zero-filled). Returns the aligned `(strategy, benchmark)`
96 /// value vectors, in ascending timestamp order.
97 fn align_returns(&self, a: &Returns, b: &Returns) -> (Vec<f64>, Vec<f64>) {
98 let a_daily = self.downsample_to_daily_bins(a);
99 let b_daily = self.downsample_to_daily_bins(b);
100
101 let mut aligned_a = Vec::new();
102 let mut aligned_b = Vec::new();
103
104 for (timestamp, &a_value) in &a_daily {
105 if let Some(&b_value) = b_daily.get(timestamp) {
106 aligned_a.push(a_value);
107 aligned_b.push(b_value);
108 }
109 }
110
111 (aligned_a, aligned_b)
112 }
113
114 /// Validates that returns data is not empty.
115 fn check_valid_returns(&self, returns: &Returns) -> bool {
116 !returns.is_empty()
117 }
118
119 /// Downsamples high-frequency returns to daily bins by geometric compounding.
120 ///
121 /// Within each UTC day, returns are combined via `(1 + r1)(1 + r2) - 1` to produce
122 /// the day's effective return, which is the standard convention for chaining
123 /// arithmetic period returns. For daily-frequency inputs (one return per day) the
124 /// bin value is identical to the input value, so callers that already operate on
125 /// daily returns observe no behavior change.
126 fn downsample_to_daily_bins(&self, returns: &Returns) -> Returns {
127 let nanos_per_day = 86_400_000_000_000; // Number of nanoseconds in a day
128 let mut daily_bins = BTreeMap::new();
129
130 for (×tamp, &value) in returns {
131 // Calculate the start of the day in nanoseconds for the given timestamp
132 let day_start = timestamp - (timestamp.as_u64() % nanos_per_day);
133
134 // Geometrically compound returns within each day
135 let entry = daily_bins.entry(day_start).or_insert(0.0_f64);
136 *entry = (1.0_f64 + *entry).mul_add(1.0_f64 + value, -1.0_f64);
137 }
138
139 daily_bins
140 }
141
142 /// Calculates the standard deviation of returns with Bessel's correction.
143 fn calculate_std(&self, returns: &Returns) -> f64 {
144 let n = returns.len() as f64;
145 if n < 2.0 {
146 return f64::NAN;
147 }
148
149 let mean = returns.values().sum::<f64>() / n;
150
151 let variance = returns.values().map(|x| (x - mean).powi(2)).sum::<f64>() / (n - 1.0);
152
153 variance.sqrt()
154 }
155}
156
157#[cfg(test)]
158mod tests {
159 use nautilus_core::{UnixNanos, approx_eq};
160 use rstest::rstest;
161
162 use super::*;
163
164 #[derive(Debug)]
165 struct DummyStat;
166
167 impl PortfolioStatistic for DummyStat {
168 type Item = f64;
169
170 fn name(&self) -> String {
171 "DummyStat".to_string()
172 }
173 }
174
175 const NANOS_PER_DAY: u64 = 86_400_000_000_000;
176 const BASE_NS: u64 = 1_600_000_000_000_000_000;
177
178 #[rstest]
179 fn test_downsample_compounds_intraday_returns() {
180 // Two intraday returns in the same UTC day: +5% then -5%.
181 // arithmetic sum: 0.05 + (-0.05) = 0.00 (incorrect)
182 // geometric chain: (1.05)(0.95) - 1 = -0.0025 (correct)
183 let stat = DummyStat;
184 let mut returns: Returns = BTreeMap::new();
185 returns.insert(UnixNanos::from(BASE_NS), 0.05);
186 returns.insert(UnixNanos::from(BASE_NS + 3_600_000_000_000), -0.05);
187
188 let daily = stat.downsample_to_daily_bins(&returns);
189
190 assert_eq!(daily.len(), 1);
191 let value = *daily.values().next().unwrap();
192 assert!(approx_eq!(f64, value, -0.0025, epsilon = 1e-12));
193 }
194
195 #[rstest]
196 fn test_downsample_daily_inputs_unchanged() {
197 // For one-return-per-day inputs the bin value equals the input return,
198 // so existing callers that already pass daily returns see no change.
199 let stat = DummyStat;
200 let mut returns: Returns = BTreeMap::new();
201 returns.insert(UnixNanos::from(BASE_NS), 0.01);
202 returns.insert(UnixNanos::from(BASE_NS + NANOS_PER_DAY), -0.02);
203 returns.insert(UnixNanos::from(BASE_NS + 2 * NANOS_PER_DAY), 0.015);
204
205 let daily = stat.downsample_to_daily_bins(&returns);
206
207 let values: Vec<f64> = daily.values().copied().collect();
208 assert_eq!(values.len(), 3);
209 assert!(approx_eq!(f64, values[0], 0.01, epsilon = 1e-15));
210 assert!(approx_eq!(f64, values[1], -0.02, epsilon = 1e-15));
211 assert!(approx_eq!(f64, values[2], 0.015, epsilon = 1e-15));
212 }
213
214 #[rstest]
215 fn test_downsample_chains_three_intraday_returns() {
216 // Three returns in the same day: +1%, +2%, -1%.
217 // geometric chain: (1.01)(1.02)(0.99) - 1 = 0.019998
218 let stat = DummyStat;
219 let mut returns: Returns = BTreeMap::new();
220 returns.insert(UnixNanos::from(BASE_NS), 0.01);
221 returns.insert(UnixNanos::from(BASE_NS + 3_600_000_000_000), 0.02);
222 returns.insert(UnixNanos::from(BASE_NS + 7_200_000_000_000), -0.01);
223
224 let daily = stat.downsample_to_daily_bins(&returns);
225
226 assert_eq!(daily.len(), 1);
227 let value = *daily.values().next().unwrap();
228 let expected = 1.01_f64 * 1.02 * 0.99 - 1.0;
229 assert!(approx_eq!(f64, value, expected, epsilon = 1e-12));
230 }
231}