I adapted my models for NFP using ADP and the prior NFP as features. The models came out pretty bad to be honest, with a high error on each. Still, here's the shot in the dark prediction: image I'll see what else I can add as features next month ...

Diary of a Pro Quant

I added a new feature, the prior ISM PMI. That seemed to help the models quite a bit: image Thiel-Sen was only off by .5 this time, still with no hyperparameter tuning.

Diary of a Pro Quant

The data came in at 46.0. This was way below expectations and its fair to say my models did poorly. Thiel-Sen came the closest, but still wasn't very close. I'll tune them and try again next month!

Diary of a Pro Quant

ISM Manufacturing PMI is due out at 10:00AM EST and the FF calendar forecasts it to be 47.2. My models suggest it may come in higher. I trained 5 separate machine learning models on regional PMI data for June and they all produced higher estimates ...

Cable Update (GBPUSD)

ISM Manufacturing PMI is due out at 10:00AM EST and the FF calendar forecasts it to be 47.2. My models suggest it may come in higher. I trained 5 separate machine learning models on regional PMI data for June and they all produced higher estimates ...

USD/JPY Discussion

ISM Manufacturing PMI is due out at 10:00AM EST and the FF calendar forecasts it to be 47.2. My models suggest it may come in higher. I trained 5 separate machine learning models on regional PMI data for June and they all produced higher estimates ...

EURUSD

Using historical data on regional Manufacturing PMI data, I created 5 basic machine learning models to predict ISM Manufacturing PMI. Per the ForexFactory.com Calendar, the market expects the data to come in at 47.2. The mean forecast my models ...

Diary of a Pro Quant

July has been the second strongest month for GBP/USD historically: image However, the first part of the month has been weak. image

Cable Update (GBPUSD)

Historically, July has been the best month for AUD/USD: image

AUD/USD

Seasonal update for July: image Historically it's been a lackluster month for EUR/USD.

EURUSD

It turns out that encoding didn't make much difference. I tried a few things out and the accuracy ranged from 85% to 87%, barely any different from my starting point of 86%. On to cross-validation.

Diary of a Pro Quant

This sounds like a martingale strategy and won't work longer term.

Have a quick question about Fundamental Analysis and Grid Trading

I created some very simple ML models using logistic regression and extreme gradient boosting to predict the 1 day future return of the S&P 500. I didn't try to make anything useful quite yet, it's more just getting it to work. The logistic ...

Diary of a Pro Quant

I doubt the statement itself does much, but I'm eager to see the dot plot. When I checked last the odds still favor a July hike so any insight into that will move the market.

Clemmo's Trading Odyssey

Sorry for the brief hiatus here, I'm working on an ML model at the moment. I'm actually newer to time series forecasting with machine learning, but based on what I've read XGBoost looks like the way to go. I'm trying to get something basic up and ...

Diary of a Pro Quant

Very interesting! I prefer Java to Python, but finance has (mostly) adopted Python so I have too. You're right, there's a few different ways of doing this and it's not clear which approach best. This beta version of cyclical charting is low ...

Diary of a Pro Quant

Your cumulative return looks almost the same as mine when I used median. One slight difference is yours dips at the very end, and mine doesn't because I'm excluding the last two trading days. The reason for that is some years have more days than ...

Diary of a Pro Quant