Performance of the passive aggressive mean reversion trading algorithm with survivorship-bias free data on the Helsinki Stock Exchange
Toikka, Toivo (2018)
Kandidaatintutkielma
2018
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2018060125131
https://urn.fi/URN:NBN:fi-fe2018060125131
Tiivistelmä
In this thesis, we are studying Performance of the Passive Aggressive Mean Reversion (PAMR) trading algorithm with survivorship-bias-free data from the Helsinki Stock Exchange. This algorithm uses passive aggressive online learning algorithm and mean-reversion trading strategy to its advantage. We are interested could we accept this strategy as a viable trading strategy and how survivorship biased data affects to results.
PAMR’s performance was simulated with two datasets. One dataset contained historical close prices of companies from OMXH25-index in 2008 and another dataset contained historical close prices of companies from OMXH25-index in 2018. The study period was 12.03.2008 - 09.03.2018.
The algorithm performed well when comparing to OMXH25-index, but survivorship biased data gave better results than survivorship bias-free data. From this, we can make a conclusion that when using survivorship biased data results are overly confident in the study period.
PAMR’s performance was simulated with two datasets. One dataset contained historical close prices of companies from OMXH25-index in 2008 and another dataset contained historical close prices of companies from OMXH25-index in 2018. The study period was 12.03.2008 - 09.03.2018.
The algorithm performed well when comparing to OMXH25-index, but survivorship biased data gave better results than survivorship bias-free data. From this, we can make a conclusion that when using survivorship biased data results are overly confident in the study period.