Everything I build with actuals and higher volume stocks (mid to big caps) data is going through the roof (out of sample, short timeframe (about a year), but significant. The backtest is a bit worse than the OOS time, so I suspect the data provided gets better…
Anybody build systems with actuals data making good experieces?
I really like the actuals data, it is more accurate than the previous historical data. I did not get the level of boost you experienced, but it is better
This may be due to the recent popularity of actual data-related factors.
But at least it proves that the actual data may be useful and is unlikely to be seriously biased.
well, I got different stuff in my head concerning actuals data
Popularity
more companies posting actuals now than in the past
recent data (that comes into the database from the day p123 implemented might be better than the historic (that was bulk loaded when P123 implemented it)
surprise effect --> just look at the breakout of SMCI (the one beginning of the year, they posted actuals and the stock went up big) --> leads to character change of the price behaviour, very powerfull edge (also comes into play when Eearnings are reported (PEAD), but everybody has earnings data nowdays, so everybody fights for this edge)
so Post Actuals announcement drift (PAAD ;-)) might be more powerfull than origninal PEAD...
Very, very interesting, it works better with mid an big caps in my backtests, small caps nor working well...
Still a bit shy to allocate big money to it, but watching OOS!!!
One more thing --> all my actuals models explode 2020 --> that was the change to Factset, so they might have better actuals data...
Is it possible that this is because P123 collects its own actual data?
fed pumped the market with money after April 2020, price went up for everything, run the same on data from 2014 to 2024, if the strategy is decent it should return around 8x to 11x from jan 2014 to around jan 2020 on major tech,
regarding data P123 keep tinkering and this sometimes is reflected when you see returns, one week something next week nada, anyhow I keep with the platform because I made $$$ regardless.
Yes, that could be it (FED Pumping, Market structure definitly changed then!)
Results from 2014 - 2020 are about 30% / 35% DD, so very decent for a mid to big cap model.
The actuals models I am looking at are out of sample since about a year and they did great, so I will start put some allocation into them (10-20%).
I am sorry, i have been reading two threads on this topic but is unclear to me what actuals are. Do you mean actual data released by companies as opposed to expected ?
Do you mind providing an example ?
Thank you
There is some background here:
