Picking daily direction?
- Thread starter Seb_hedge
- Start date
That’s a nice exercise
Checkout linda raschke
The best you could do is to get your hands dirty by testing hypotheses over data in excel or whatever.
What if scenarios.
Maybe the open relative to the previous close ?
Maybe the open relative to the previous day HL ?
Maybe the overnight range relative the open ?
Maybe we have a trend over the previous days ?
…
A reasonable level of certainty might be 60%
Maybe most of the time we don’t know …
But once in a while we might have 80% probability.
Checkout linda raschke
The best you could do is to get your hands dirty by testing hypotheses over data in excel or whatever.
What if scenarios.
Maybe the open relative to the previous close ?
Maybe the open relative to the previous day HL ?
Maybe the overnight range relative the open ?
Maybe we have a trend over the previous days ?
…
A reasonable level of certainty might be 60%
Maybe most of the time we don’t know …
But once in a while we might have 80% probability.
That's like asking for the Holy Grail or the Keys to the Kingdom. I imagine if anyone holds such knowledge he's not going to share it publicly.
If you were to pursue this path, I'd suggest first compiling statistics and metrics of your chosen market and see what answers you can find through testing that data. If you're a decent programmer you should be able to get somewhere within a few months time if you're very dedicated and do it full time, but more likely it will take you at least a year. If you're not a programmer (like me), expect it to take you years.
If you were to pursue this path, I'd suggest first compiling statistics and metrics of your chosen market and see what answers you can find through testing that data. If you're a decent programmer you should be able to get somewhere within a few months time if you're very dedicated and do it full time, but more likely it will take you at least a year. If you're not a programmer (like me), expect it to take you years.
%%
Exactly;
much of it depends on the market+ time frame. SEPT may differ.
REAL estate tends to uptrend well\ unless in a high crime area.
Just get Daily data into excel and check the PnL
If this happens then what if I buy the open & sell the close ?
If the inverse happens then what if I sell the open & buy the close ?
Looks like on the SPX (1 Year of Data)
Buying the open (if yesterday's (high-open)/(high-open+open-low) < 0.4) and selling the close is profitable.
Selling the open (if yesterday's (high-open)/(high-open+open-low) > 0.6) and buying the close is profitable.
Buying the open (if open > than previous close) and selling the close is profitable.
Selling the open (if open < than previous close) and buying the close is profitable.
Better not fade them
Try some conditions and see for yourself.
Here we're buying the open (if previous day close > previous day open) and selling the close
Here we're selling the open (if previous day close < previous day open) and buying the close
Better fade them
Here we're combining B+C but it doesn't beat B alone
Here
We're buying the open (if open > previous high) and selling the close
We're selling the open (if open < previous low) and buying the close
Better not fade them.
What's funny ...
If you buy the open and sell the close
When the open is above the previous close
You would have made money only 25% of the time.
If you sell the open and buy the close
When the open is below the previous close
You would have made money only 31% of the time.
But we've seen it's been a profitable strategy.
On paper ... Without costs nor potential margin calls.
8pts in average for the long strategy (min -107, max 96)
4pts in average for the short strategy (min -149, max 141).
Inversely ...
You can lose money in average on sure things.
Like betting on the favorites in sport betting.
You might win 80% of the time,
But lose money overall !
If this happens then what if I buy the open & sell the close ?
If the inverse happens then what if I sell the open & buy the close ?
Looks like on the SPX (1 Year of Data)
Buying the open (if yesterday's (high-open)/(high-open+open-low) < 0.4) and selling the close is profitable.
Selling the open (if yesterday's (high-open)/(high-open+open-low) > 0.6) and buying the close is profitable.
Buying the open (if open > than previous close) and selling the close is profitable.
Selling the open (if open < than previous close) and buying the close is profitable.
Better not fade them
Try some conditions and see for yourself.
Here we're buying the open (if previous day close > previous day open) and selling the close
Here we're selling the open (if previous day close < previous day open) and buying the close
Better fade them
Here we're combining B+C but it doesn't beat B alone
Here
We're buying the open (if open > previous high) and selling the close
We're selling the open (if open < previous low) and buying the close
Better not fade them.
What's funny ...
If you buy the open and sell the close
When the open is above the previous close
You would have made money only 25% of the time.
If you sell the open and buy the close
When the open is below the previous close
You would have made money only 31% of the time.
But we've seen it's been a profitable strategy.
On paper ... Without costs nor potential margin calls.
8pts in average for the long strategy (min -107, max 96)
4pts in average for the short strategy (min -149, max 141).
Inversely ...
You can lose money in average on sure things.
Like betting on the favorites in sport betting.
You might win 80% of the time,
But lose money overall !
Last edited:
This statement is wrong ... Had the conditional probability wrong.
Here are more accurate statistics based on SPY over 1000 days (Back into 2018)
Worth noting statistics are ...
Knowing that we opened above the previous close
There is a 77% likelihood that we're going to make a new high (G|F)
There is a 24% likelihood that we're going to make a new low (H|F)
The probability that we close above the new high is 65% (P|G)
Knowing that we opened below the previous close
There is a 27% likelihood that we're going to make a new high (G|-F)
There is a 70% likelihood that we're going to make a new low (H|-F)
The probability that we close below the new low is 55% (Q|H)
Need to tinker to find strong inferences.
Maybe we can check what happens for combinations like
Previous open > Previous close > open
Over 10 years worth of SPY we have these statistics for the day to come
Close > Open: 53% (meaningless)
a. New High: 12%
Close above previous high given .a: 64% (meaningless)
b. New Low: 86%
Close below previous low given .b: 58% (meaningless)
If SPY opens lower then it might be worth selling the open
Stop the loss @ previous high and take profit @ previous low.
Check if the reward to risk makes sense given the probabilities
The R:R should be at least 0.14 to 1 ...
If we can sell the previous close there's a 0.6R
39% return on risk : 0.86*0.6-0.12
Previous open > Previous close > open
Over 10 years worth of SPY we have these statistics for the day to come
Close > Open: 53% (meaningless)
a. New High: 12%
Close above previous high given .a: 64% (meaningless)
b. New Low: 86%
Close below previous low given .b: 58% (meaningless)
If SPY opens lower then it might be worth selling the open
Stop the loss @ previous high and take profit @ previous low.
Check if the reward to risk makes sense given the probabilities
The R:R should be at least 0.14 to 1 ...
If we can sell the previous close there's a 0.6R
39% return on risk : 0.86*0.6-0.12
Last edited:
I only focus on what has happened past few minutes to the past few hours.
Then react to that.
Am I keen to know the direction for the next day?
Absolutely not interested.
The more you try to predict,
the less edge you have.
Last edited: