f(x) News Trading

Hello FF! Welcome to my “f(x) News Trading” thread!

This first post will try and explain how to trade the news in the system's most basic terms but as the thread goes on we will be deep diving into a decent amount of technical concepts that will hopefully teach you something you don’t already know.

Please makes sure to read everything below before asking questions, but if anything is unclear or needs additional explanation let me know and I'll post a follow-up.

To keep things simple let’s set up some basic rules to keep the thread and discussion on track.

The system is based on 3 “indicators” and they are:

1. Price (OHLC Data)
2. 200 MA
3. News Volatility/Volatility in general.

The most important of the three is #3 since our opportunity to make money is directly related to the amount of volatility we are able to trade.

Starting off we will only be looking at the following USD pairs:

1. EUR/USD
2. GBP/USD
3. AUD/USD
4. NZD/USD
5. USD/CAD
6. USD/CHF
7. USD/JPY

These are typically the most liquid currency pairs. They will allow us to trade all of the major currencies' high-impact news events, giving us ample opportunities to make profits.

Now that the general rules are covered, I want to touch briefly on the core idea behind this strategy. It can be summed up in the paragraph below:

It’s better to focus on managing exposure than it is to focus on forecasting. By focusing on your f(x), you can know little to nothing about x and still be successful.

I touched on this in a bit more detail in the thread below so I highly recommend reading it if you haven’t already: https://www.forexfactory.com/thread/...9#post14314499

This system puts a good amount of focus on x, but the engine is our f(x).

The f(x) of this system is based on something called “dynamic replication”, a concept used by market makers to hedge non-linear payoffs using linear products.

We will do a deep dive into this concept and why it matters later but all you need to understand about it is this.

Your current position size in a trade is equal to your max position size, multiplied by the current probability of your trade being a winner. I'll detail in a follow-up post a few different ways you can estimate this probability but for some quick back-of-the-napkin math, when you enter, you are at 50%, and every couple of pips in either direction increases or decreases it by a few points. You can't constantly adjust the position so I usually do it in increments of 5-10% changes in the probability of the trade being a winner.

As long as you maintain this position sizing, you will be structuring your f(x) so that it changes from a linear to a non-linear payoff. If you can maintain a non-linear payoff, you don’t need to worry about forecast x in a directional sense. You only need to make sure you trade x when it is volatile.

Let’s wrap this first post up with somewhat of an “example” setup looking at the next high-impact news event.

Next on the docket is Canada’s CPI numbers. Will they be better or worse than expected? Who knows, and who really cares? All I care about is the volatility the event will bring.

Since this is a CAD event, we will be looking to trade USD/CAD.

Let’s look at how the price's typical volatility stacks up to this event's historical volatility.

Below shows the average open-close and high-low range for any 4-hour period of USD/CAD in addition to the average ranges for 4-hour periods that contain this event 5 minutes after the period begins.



We can see that this event does bring above-average volatility and we can expect that the range we will see on Tuesday from 7:25 am to 11:25 am is likely to have 2x or more volatility that the average 4-hour period.

Since we need volatility to take advantage of our f(x), this will be the perfect time to trade. I'll likely be looking to trade in the direction of the 200MA. I won't be entering until about 5 minutes before the event though.

Overall the long haul, your forecasting of direction will converge towards 50%. We are not too worried about x though as f(x) is our engine. All we care about x is that it's volatile as we are trying to take advantage of excess kurtosis.

Something to note about this event, most of the whipsaw (high-low range) occurs in the first 5 minutes (the chart below is for the first 5 minutes after the event). After that, the direction should be clear and the price will likely continue on the path it's currently on.

Thank you.

I love your way of thinking. Very clear and logical... Interesting to read your mathematical point of view... Will follow with interest. Thank you for sharing.

Interesting, looking forward for a more detailed post on how we implement it in real markets.
Thanks

subscribed

So far nothing but chop on the CAD inflation. Have an about a 15 pip high low range currently on USDCAD. Over next 3 hours and 45 mins there is at least a 50% chance we see a move of 45+ pips in either direction beyond the current high-low range. I am playing long atm but we’ll see how it works oitZ

Has basically played out exactly as described. we broke back above the initial range and have continued to trend in one direction for the rest of the 4-hour period

If you were playing long side gj, if you played short side, don't fret, your forecasting will converge to 50/50 over time so you were "wrong" in playing the other side. The key is to pick a side, stick to it, and dynamically manage the position.

The next event on the docket is USD news tomorrow. Will post what I am looking at for that.

Next 2 events I'll be watching:



Will most likely trade GBP/USD for both of them.

For GBP CPI our 5-minute High-Low range is about 21 pips compared to a typical average of 8. The 4-hour one is 65 pips compared to 45. So roughly 33% of the move will occur in the first 5 minutes, then the remaining 235 minutes, the range should double, most likely in one direction.


Then we have the fed fund rate. This one brings massive volatility. The 5-minute high-low range on GU is 50 pips compared to a typical average of 6.58. The 4-hour one is 114 compared to 45. So roughly 1/2 of the move will happen in the first 5 minutes and you get the point for the rest of the period.

Either way, this should provide some great volatility to trade.

You should publish a ”checklist” on how you calculate all this and how you adapt it live. Also how you manage a loosing/profitable trade.

Thanks for the feedback! I will try to post something like this today.

Look at the round-trip from GU this morning:



As expected, we saw more than 1/3 of the move happen almost instantly as the range expanded upwards for the remaining 235 minutes (7:00-11:00). Surprised how it completely turned around after 11:00. Gives a good hint at what the market is expecting for USD news. They either don't know what is happening and wanted to "re-balance" back to even or they think the news will be good for USD and they took the spike as an opportunity to get short. I'll probably be looking to trade USD news short.

Looks like long was the play on GU for fed fund rate. Have seen 50 pip range in first 2 minutes. Expect it to increase another 50+ pips for nest 235 minutes. Will see which direction it grows in...

I was playing short so am mostly reduced to only 20% of my position size. Will keep scaling out if it keeps running higher, otherwise will scale in further if it reverses.

Hi Riskcuit!
Any chart-examples of your trades?
How are you ”scaling” the positions?
If you where short, how do you manage this loosing position ?

Best regards

Here is some great Saturday reading I’d recommend to anyone looking for a clearer understanding of dynamic replication:

https://www.glynholton.com/notes/dynamic_hedging/

https://cims.nyu.edu/~ritter/kolm2019dynamic.pdf

I will try and post more chart examples this weekend if I have time.

Both of these talk about dynamic replication in the framework of hedging an option using the underlying. We are not hedging or using options, but we are trying to relocate the non-linear payoff of a long option.

these payoffs are +EV when volatility increases. This is why we try to replicate them Trading news, because as the graphs I’ve posted for a couple events show, volatility drastically increases during these windows compared to what they typically are.

This is very interesting.

Do you automate the exposure changing or is it a manual process?

thank you subscribed

I typically do it manually. Sometimes I will automate it if I need to step away but I typically don't like to leave trades on if I can't watch them.

Here is some data for NFP tomorrow on GBPUSD.



The average High-Low range for a 5-minute candle is about 9 pips. For NFP, the first 5 minutes after the event typically produce a range of 38.66+ pips.



The average 4-hour High-Low range for a 4-hour candle is 36 pips. For NFP, the first 4 hours after the event typically produce a range of 80 pips+.


So we can expect to see the first 1/2 of the range occur in the first 5 minutes of the event, then we should see the range double in size (in one direction or another) over the remaining 235 minutes. Trade accordingly!

Hello Riskcuit, it's interesting to watch sentiment indicator on myfxbook or elsewhere, since let's say you see downtrend on USD/CAD but 78% people are buying, you can expect price to continue downwards after the news. In this example I would wait until price makes move up (most people will want to buy quickly) but you will go sell. It happens 90% of the time.

Let me know what you think. Maybe you can check it from history.

SF

unless one scalps lower timeframes price 95% of the time returns to the point of the news release so i found it easier and less risky to trade the reversion back to the news announcement...for example im short oil and gold since price will fall to the saudi news event

I decided to register because I'm interested in collaborating with you on this. I operate a multi-strategy quantitative portfolio and I think a strategy along these lines would add value, i.e. improve the overall portfolio sharpe.

I have a few questions about your strategy.

1.) How are you measuring historical volatility?

You could be scraping the forexfactory calendar, converting the dates into something datetime library compatible (if you're using Python), resampling the timeframe to intraday, then measuring historical volatility that way. Its a lot of work though so I wonder if you're using something more developer friendly like a Bloomberg Terminal instead? For that matter, are you doing this in Python, Excel, etc?

Furthermore, the impact of these fundamental events changes over time. CPI, for example, is highly important now as it directly impacts monetary policy which drives FX. However, it wasn't so important a few years ago when inflation wasn't a problem. So I'm wondering what you're doing to account for the inherently dynamic relationship between fundamentals and the market. Are you using a narrow lookback period or perhaps weighting more recently events heavier?

2.) I'm not clear on how you're achieving a non-linear payoff using a linear instrument, can you provide an example?

Something like stop-limit orders might make this quasi-achievable, but I think I'm missing something on your overall approach. Typically this would be done using something like a strangle options strategy.

3.) It looks like a core component of your strategy is estimating the probability of your trade being a winner. Are you, or have you considered, using options greek math to compute this? I don't think it would be terribly difficult to do this using a modified computation of delta.