AlgorithmSpeedTracker
Bases: Object
Accumulates periodic samples of the engine's speed counters (data points processed, history data points, backtest days processed) while a backtest runs, and computes the throughput and progress metrics consumed by the in-run Analyses.AlgorithmSpeedAnalysis. All rates are computed between samples, so setup time before the first sample is excluded.
RECENT_WINDOW_SAMPLES: int = 5
The number of trailing samples that make up the "recent" window used by the windowed rates. At the ~30 second in-run analysis cadence this spans roughly the last two minutes.
sample_count: int
The number of samples recorded so far.
total_days: int
The total number of calendar days the backtest will run.
processed_days: int
The number of calendar days the backtest has processed as of the latest sample.
progress: float
The backtest progress as of the latest sample, in the <0, 1> range.
elapsed: timedelta
The wall-clock time elapsed since the backtest started, as of the latest sample.
sampled_span: timedelta
The wall-clock time between the first and the latest sample, that is, the period the rates are measured over.
has_data_point_counts: bool
Whether the main loop data point counter is being fed. When the counter is not wired in (it reads zero), data-point-based rates are not meaningful and should not be used.
data_points_per_second: Optional[float]
The average data points processed per second over the whole sampled span, including history data points to match the speed the engine reports on completion. Null when there are not enough samples to measure.
initial_data_points_per_second: Optional[float]
The average data points processed per second over the first RECENT_WINDOW_SAMPLES samples, used as the early-run baseline for degradation detection. Null when there are not enough samples to measure.
add_sample(sample: AlgorithmSpeedSample) -> None
Records a sample of the cumulative speed counters. Samples with a non-increasing elapsed time are ignored so rates are always computed over positive time deltas.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
sample
|
AlgorithmSpeedSample
|
The sample of the cumulative speed counters. |
required |
estimated_remaining_time(
skip_last: int = 0,
) -> Optional[timedelta]
The estimated wall-clock time left for the backtest to complete, projecting the recent calendar-days-per-second pace over the remaining backtest days.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
skip_last
|
int
|
Number of trailing samples to skip, to evaluate the projection as of a previous run. |
0
|
Returns:
| Type | Description |
|---|---|
Optional[timedelta]
|
The estimate, zero when the backtest already reached its end date, or null when the recent pace is zero or there are not enough samples to measure. |
recent_data_points_per_second(
skip_last: int = 0,
) -> Optional[float]
The average data points processed per second over the recent window, including history data points.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
skip_last
|
int
|
Number of trailing samples to skip, to evaluate the window as of a previous run. |
0
|
Returns:
| Type | Description |
|---|---|
Optional[float]
|
The windowed rate, or null when there are not enough samples to measure. |
recent_days_per_second(
skip_last: int = 0,
) -> Optional[float]
The average backtest calendar days processed per wall-clock second over the recent window.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
skip_last
|
int
|
Number of trailing samples to skip, to evaluate the window as of a previous run. |
0
|
Returns:
| Type | Description |
|---|---|
Optional[float]
|
The windowed rate, or null when there are not enough samples to measure. |
recent_history_data_points(skip_last: int = 0) -> int
The number of history data points served over the recent window.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
skip_last
|
int
|
Number of trailing samples to skip, to evaluate the window as of a previous run. |
0
|
recent_history_data_points_share(
skip_last: int = 0,
) -> Optional[float]
The share of the data points processed over the recent window that were served by the history provider, in the <0, 1> range.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
skip_last
|
int
|
Number of trailing samples to skip, to evaluate the window as of a previous run. |
0
|
Returns:
| Type | Description |
|---|---|
Optional[float]
|
The share, or null when there are not enough samples or no data points were processed in the window. |