TypeError: cannot do positional indexing on RangeIndex with these indexers [nan] of type float - Backtesting #9831

 freqtrade backtesting  -c config.json --timerange 20170101-20240206 --strategy Test  --freqaimodel XG
BoostRegressor --timeframe 8h
Explain the problem you have encountered
2024-02-18 18:56:05,122 - freqtrade.optimize.backtesting - INFO - Backtesting with data from 2019-09-16 00:00:00 up to 2024-02-06 00:00:00 (1604 days).
/home/the-box/Github/freqtrade/freqtrade/data/metrics.py:147: FutureWarning: The behavior of Series.idxmin with all-NA values, or any-NA and skipna=False, is deprecated. In a future version this will raise ValueError
  else max_drawdown_df['drawdown'].idxmin()
2024-02-18 18:56:08,230 - freqtrade - ERROR - Fatal exception!
Traceback (most recent call last):
  File "/home/the-box/Github/freqtrade/freqtrade/main.py", line 42, in main
    return_code = args['func'](args)
                  ^^^^^^^^^^^^^^^^^^
  File "/home/the-box/Github/freqtrade/freqtrade/commands/optimize_commands.py", line 58, in start_backtesting
    backtesting.start()
  File "/home/the-box/Github/freqtrade/freqtrade/optimize/backtesting.py", line 1397, in start
    results = generate_backtest_stats(
              ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/the-box/Github/freqtrade/freqtrade/optimize/optimize_reports/optimize_reports.py", line 559, in generate_backtest_stats
    strat_stats = generate_strategy_stats(pairlist, strategy, content,
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/the-box/Github/freqtrade/freqtrade/optimize/optimize_reports/optimize_reports.py", line 438, in generate_strategy_stats
    'calmar': calculate_calmar(results, min_date, max_date, start_balance),
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/the-box/Github/freqtrade/freqtrade/data/metrics.py", line 300, in calculate_calmar
    _, _, _, _, _, max_drawdown = calculate_max_drawdown(
                                  ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/the-box/Github/freqtrade/freqtrade/data/metrics.py", line 150, in calculate_max_drawdown
    high_date = profit_results.loc[max_drawdown_df.iloc[:idxmin]['high_value'].idxmax(), date_col]
                                   ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^
  File "/home/the-box/Github/freqtrade/.venv/lib/python3.11/site-packages/pandas/core/indexing.py", line 1153, in __getitem__
    return self._getitem_axis(maybe_callable, axis=axis)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/the-box/Github/freqtrade/.venv/lib/python3.11/site-packages/pandas/core/indexing.py", line 1691, in _getitem_axis
    return self._get_slice_axis(key, axis=axis)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/the-box/Github/freqtrade/.venv/lib/python3.11/site-packages/pandas/core/indexing.py", line 1726, in _get_slice_axis
    labels._validate_positional_slice(slice_obj)
  File "/home/the-box/Github/freqtrade/.venv/lib/python3.11/site-packages/pandas/core/indexes/base.py", line 4197, in _validate_positional_slice
    self._validate_indexer("positional", key.stop, "iloc")
  File "/home/the-box/Github/freqtrade/.venv/lib/python3.11/site-packages/pandas/core/indexes/base.py", line 6637, in _validate_indexer
    self._raise_invalid_indexer(form, key)
  File "/home/the-box/Github/freqtrade/.venv/lib/python3.11/site-packages/pandas/core/indexes/base.py", line 4306, in _raise_invalid_indexer
    raise TypeError(msg)
TypeError: cannot do positional indexing on RangeIndex with these indexers [nan] of type float
Steps to reproduce:
Run the commands above with my strategy

The weird thing is that this happens when I run on 8h, running on 12h and 1d works perfectly fine.
Observed Results:
What happened?

The error above
What did you expect to happen?

Backtest to complete normally
I also ran the freqtrade lookahead-analysis to see if I was doing something weird, but it passed the test
I make my indicators, I don't think I'm doing anything wrong with this (below snippet).

I use it to calculate an average based on a rolling period of previous candles, so for each new candle, look back to start_at_bar and compute something
            for _index, (_, _) in enumerate(dataframe.iterrows()):
                for i in range(start_at_bar + 1):
                    y = dataframe['ohlc4'].shift(i).iloc[_index]
                    # Do some arithmetic, no more .shift() nor .iloc[i]




    

Update: I tried running it on dry mode to see if the logs showed something
2024-02-18 20:03:14,199 - freqtrade.strategy.interface - WARNING - Unable to analyze candle (OHLCV) data for pair BTC/USDT:USDT: Dataframe returned from strategy has mismatching last close price.
I'm not sure if this is the reason
    heikinashi = qtpylib.heikinashi(dataframe)
    dataframe["open"] = heikinashi["open"]
    dataframe["close"] = heikinashi["close"]
    dataframe["high"] = heikinashi["high"]
    dataframe["low"] = heikinashi["low"]
Is this allowed?
Reverting to the normal candles (removing the snippet above) causes the same error to appear on backtesting on 1d
However, this happens if I run the historical data, if I run the last few months, the backtest is successful .
Maybe something with this?
FutureWarning: The behavior of Series.idxmin with all-NA values, or any-NA and skipna=False, is deprecated. In a future version this will raise ValueError
  else max_drawdown_df['drawdown'].idxmin()
2024-02-18 20:19:19,329 - freqtrade - ERROR - Fatal exception!