To see what the impact of the market trend has on the Bollinger
Band model we can look at the results of 2018 and 2019. Bitcoin started 2018
started with a BTC price of $13,850 and ended at $3,747. It dropped rapidly
early and was generally had a downward move. In 2019 it started at $3,747 and
was at $8,327 on September 30th. Before running the simulation for
each year separately, we narrowed the parameters ranges. We used 7 to 16 days
and standard deviation multipliers of 1.7, 1.9, and 2.1. Following are the top
10 combinations from each year (2018 on left, 2019 on right).
|
Net Worth
|
Days Desired
|
Std Dev Multiplier
|
#Trades
|
# Buys
|
# Sells
|
Net Worth
|
Days Desired
|
Std Dev Multiplier
|
#Trades
|
# Buys
|
# Sells
|
|
|
$6,716
|
15
|
1.9
|
111
|
60
|
51
|
$13,764
|
7
|
2.1
|
98
|
49
|
49
|
|
|
$6,548
|
14
|
1.7
|
134
|
72
|
62
|
$13,379
|
7
|
1.7
|
124
|
62
|
62
|
|
|
$6,547
|
13
|
1.7
|
140
|
74
|
66
|
$12,861
|
7
|
1.9
|
110
|
55
|
55
|
|
|
$6,352
|
15
|
1.7
|
126
|
67
|
59
|
$12,714
|
8
|
2.1
|
90
|
44
|
46
|
|
|
$6,323
|
14
|
1.9
|
109
|
59
|
50
|
$12,424
|
9
|
1.7
|
111
|
55
|
56
|
|
|
$6,190
|
12
|
1.7
|
137
|
72
|
65
|
$12,393
|
8
|
1.9
|
100
|
49
|
51
|
|
|
$6,054
|
14
|
2.1
|
95
|
52
|
43
|
$12,364
|
11
|
1.7
|
97
|
48
|
49
|
|
|
$6,050
|
16
|
1.9
|
107
|
57
|
50
|
$12,330
|
10
|
1.7
|
103
|
51
|
52
|
|
|
$6,008
|
16
|
2.1
|
89
|
49
|
40
|
$12,223
|
8
|
1.7
|
112
|
55
|
57
|
|
|
$5,997
|
16
|
1.7
|
125
|
65
|
60
|
$12,153
|
16
|
1.7
|
78
|
38
|
40
|
The top net worth combination of Bollinger Bands parameters was
quite different for the 2 years. The 2018 (downward trend) parameters were
closer to the combinations for the 2 years together. That makes sense because
the BTC price on September 30th, 2019 was lower than the opening
price on January 1st, 2018.
Unfortunately, the only profitable combination occurred in a
rising market. Furthermore, a buy and hold strategy in 2019 would have a net
profit of $12,223. That’s going to be a hard number to beat. But we have one more
idea. We’re going to use some of the previous day’s prices and use both linear
and quadratic regression to identify a recent trend. This will add a 3rd
parameter and complicate things.
This might take a while. I’ll use Twitter, @billlanke, to indicate
when I post some results.
No comments:
Post a Comment