Surprising expectations

98 views
Skip to first unread message

Henrik in Oslo

unread,
Nov 11, 2013, 7:19:56 PM11/11/13
to mnemosyne-...@googlegroups.com
I run Mnem 2.2.1 on Mac OS X10.8.5
I've turned on the "Card Statistics" plugin and notice

1. Grades 2,3,4 hardly make a difference to  Next Expected Repetition day
2. Grade 5 is predicted to result in shorter time until repetition than grade 4, 3

(the attached snap-shots are taken while I was learning one day ahead of schedule)

Any explanations?
/ Henrik
grade_days_01.png
grade_days_02.png
grade_days_03.png
grade_days_04.png

Peter Bienstman

unread,
Nov 12, 2013, 3:56:53 AM11/12/13
to mnemosyne-...@googlegroups.com
If you're reviewing ahead of time, I made some changes to the
algorithm to prevent the intervals from growing exponentially large,
especially for grade 5. So I really recommend against using review
ahead of schedule too much.

(Also I didn't write that plugin, so I can't really vouch for its
accuracy :-) )

Peter

Quoting Henrik in Oslo <001h...@gmail.com>:

> I run Mnem 2.2.1 on Mac OS X10.8.5
> I've turned on the "Card Statistics" plugin and notice
>
> 1. Grades 2,3,4 hardly make a difference to Next Expected Repetition day
> 2. Grade 5 is predicted to result in *shorter* time until repetition than
> grade 4, 3
>
> (the attached snap-shots are taken while I was learning one day ahead of
> schedule)
>
> Any explanations?
> / Henrik
>
> --
> You received this message because you are subscribed to the Google
> Groups "mnemosyne-proj-users" group.
> To unsubscribe from this group and stop receiving emails from it,
> send an email to mnemosyne-proj-u...@googlegroups.com.
> To post to this group, send email to mnemosyne-...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/mnemosyne-proj-users/1000eac1-21a7-4ae7-a36f-8fe1106f55a2%40googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.



Henrik in Oslo

unread,
Nov 16, 2013, 8:41:18 AM11/16/13
to mnemosyne-...@googlegroups.com
[Hello plugin-creator, any comments :-) ?]

Assuming the plugin  predicts fairly correctly (obviously the intended randomness distorts)

1. Previous discussion: I was reviewing 1 day ahead of schedule. The (adjusted) algorithm for the given examples results in the following: 
- Grade 2,3,4 all gives 107 days. Grade 5 gives 47
Grade 2,3,4 all gives 117 days. Grade 5 gives 45
- Grade 2,3,4 all gives 243 days. Grade 5 gives 98
- Grade 2,3,4 all gives 264 days. Grade 5 gives 102
a. Please explain why there is no differentiation between grades 2,3 and 4 (see also todays example)
b. Grade 5 resulting in (an adjusted) less than 50% of grade 3 is .... surprising. Comment?

2. Today I reviewed only todays "lesson" (nothing ahead of time)
See attachment. 
 
a. Examples A-F: Grading 2,3,4,5 hardly differentiates. Why?
b. Examples G-H: Grading 2 or 3 gives same result.  Grading 4 or 5 also gives practically same result (doubling result of 2,3). How come?

/ Henrik


On Tuesday, 12 November 2013 09:56:53 UTC+1, Peter Bienstman wrote:
If you're reviewing ahead of time, I made some changes to the  
algorithm to prevent the intervals from growing exponentially large,  
especially for grade 5. So I really recommend against using review  
ahead of schedule too much.

(Also I didn't write that plugin, so I can't really vouch for its  
accuracy :-) )

Peter

Quoting Henrik in Oslo <001h...@gmail.com>:

> I run Mnem 2.2.1 on Mac OS X10.8.5
> I've turned on the "Card Statistics" plugin and notice
>
> 1. Grades 2,3,4 hardly make a difference to  Next Expected Repetition day
> 2. Grade 5 is predicted to result in *shorter* time until repetition than
> grade 4, 3
>
> (the attached snap-shots are taken while I was learning one day ahead of
> schedule)
>
> Any explanations?
> / Henrik
>
> --
> You received this message because you are subscribed to the Google  
> Groups "mnemosyne-proj-users" group.
> To unsubscribe from this group and stop receiving emails from it,  
Grade_diff.png

Vit

unread,
Nov 16, 2013, 11:17:13 PM11/16/13
to mnemosyne-...@googlegroups.com
 Hello. Henrik.
 
I've been having this problem for 3-4mo.
I knew the culprit was not the Plugin.
 
Today, I conducted one more test.   See the Protocol below.
I tested 3 cards with and without plugin.
 
a. With Plugin  the Intervals were    6 6 6 6
b. Without Plugin:
       Card-1  I graded 4, Interval was   6
       Card-2  I graded 3,                        6
       Card-3                2                         7   ( randomness in play ).
 
 The Protocol.
                            Plugin is activated.
1. Activate Untagged cards ( there is the reason to work with Untagged cards).
2. When the first card is shown, hit "Show Answer" but do NOT rate it !!!!
    Record the Intervals for buttons 2 3 4 5.
3. Skip the card by using this 'trick':   
     Edit --- Tag it 000 --- OK  ( do not use ENTER or Ctrl_Enter ).
    The card will 'disappear' from the screen.
4. In this manner, do as many cards as you wish.
 
                Phase 2 :  Plugin is de-activated.
5. Activate cards tagged 000; 
6. Rate each card using different Buttons.
7. Go to Browser.   Compare Next Intervals with and without plugin.
 
We have food for thoughts.
 
Good day
 
Vit
 
-----------------------------------

On Saturday, November 16, 2013 8:41:18 AM UTC-5, Henrik in Oslo wrote:
[Hello plugin-creator, any comments :-) ?]

Assuming the plugin  predicts fairly correctly ....
............

Gnome

unread,
Nov 17, 2013, 3:57:38 AM11/17/13
to mnemosyne-...@googlegroups.com
Hi all. Indeed the problem seems to be with the plugin and not with mnemosyne. I did a test where 20 cards in each grade was given a corresponding grade. 20 cards where given grade 2, another 20 grade 3 and so on.
Here is the results:
Grade 2: average=1,65 days  standard deviation=0,49
Grade 3: average=3,45 days  standard deviation=0,51
Grade 4: average=4,4 days  standard deviation=0,5
Grade 5: average=7,55 days  standard deviation=0,51
I did not have the plugin installed though, so I don't know if it alters the grading algorithm in mnemosyne.

Peter Bienstman

unread,
Nov 18, 2013, 2:58:45 AM11/18/13
to mnemosyne-...@googlegroups.com
Hi,

If grade 2 and 3 give the same results, you're hitting the lower limit
of 'easiness' of 1.3 (The original SM2 algorithm also has this
built-in). In case you want the nitty-gritty, I paste some of the code
here. (As you'll understand by now, there is a very good reason
Mnemosyne does not show you the intervals it intends to schedule, as
there are too many corner cases which are confusing and would require
considerate in-depth explanation :-) )

Cheers,

Peter

if timing in ["LATE", "ON TIME"]:
if new_grade == 2:
card.easiness -= 0.16
if new_grade == 3:
card.easiness -= 0.14
if new_grade == 5:
card.easiness += 0.10
if card.easiness < 1.3:
card.easiness = 1.3
if card.ret_reps_since_lapse == 1:
new_interval = 6 * DAY
else:
if new_grade == 2 or new_grade == 3:
if timing in ["ON TIME", "EARLY"]:
new_interval = actual_interval * card.easiness
else:
# Learning late and interval was too long, so don't
# increase the interval and use scheduled_interval
# again as opposed to the much larger
# actual_interval * card.easiness.
new_interval = scheduled_interval
if new_grade == 4:
new_interval = actual_interval * card.easiness
if new_grade == 5:
if timing in ["EARLY"]:
# Learning ahead and interval was too short. To avoid
# that the intervals increase explosively
when learning
# ahead, take scheduled_interval as opposed to the
# much larger actual_interval * card.easiness.
new_interval = scheduled_interval
else:
new_interval = actual_interval * card.easiness

Quoting Henrik in Oslo <001h...@gmail.com>:

> [Hello plugin-creator, any comments :-) ?]
>
> Assuming the plugin predicts fairly correctly (obviously the intended
> randomness distorts)
>
> 1. Previous discussion: I was reviewing *1 day* ahead of schedule. The
> (adjusted) algorithm for the given examples results in the following:
> - Grade 2,3,4 all gives 107 days. Grade 5 gives 47
> - Grade 2,3,4 all gives 117 days. Grade 5 gives 45
> - Grade 2,3,4 all gives 243 days. Grade 5 gives 98
> - Grade 2,3,4 all gives 264 days. Grade 5 gives 102
> a. Please explain why there is no differentiation between grades 2,3 and 4
> (see also todays example)
> b. Grade 5 resulting in (an adjusted) less than 50% of grade 3 is ....
> surprising. Comment?
>
> 2. Today I reviewed only *todays "lesson" *(nothing ahead of time)
> See attachment.
>
> a. Examples A-F: Grading 2,3,4,5 hardly differentiates. Why?
> b. Examples G-H: Grading 2 or 3 gives same result. Grading 4 or 5 also
> gives practically same result (doubling result of 2,3). How come?
>
> / Henrik
>
>
> On Tuesday, 12 November 2013 09:56:53 UTC+1, Peter Bienstman wrote:
>>
>> If you're reviewing ahead of time, I made some changes to the
>> algorithm to prevent the intervals from growing exponentially large,
>> especially for grade 5. So I really recommend against using review
>> ahead of schedule too much.
>>
>> (Also I didn't write that plugin, so I can't really vouch for its
>> accuracy :-) )
>>
>> Peter
>>
>> Quoting Henrik in Oslo <001h...@gmail.com <javascript:>>:
>>
>> > I run Mnem 2.2.1 on Mac OS X10.8.5
>> > I've turned on the "Card Statistics" plugin and notice
>> >
>> > 1. Grades 2,3,4 hardly make a difference to Next Expected Repetition
>> day
>> > 2. Grade 5 is predicted to result in *shorter* time until repetition
>> than
>> > grade 4, 3
>> >
>> > (the attached snap-shots are taken while I was learning one day ahead of
>> > schedule)
>> >
>> > Any explanations?
>> > / Henrik
>> >
>> > --
>> > You received this message because you are subscribed to the Google
>> > Groups "mnemosyne-proj-users" group.
>> > To unsubscribe from this group and stop receiving emails from it,
>> > send an email to
>> mnemosyne-proj-u...@googlegroups.com<javascript:>.
>>
>> > To post to this group, send email to
>> mnemosyne-...@googlegroups.com<javascript:>.
>>
>> > To view this discussion on the web visit
>> >
>> https://groups.google.com/d/msgid/mnemosyne-proj-users/1000eac1-21a7-4ae7-a36f-8fe1106f55a2%40googlegroups.com.
>>
>> > For more options, visit https://groups.google.com/groups/opt_out.
>>
>>
>>
>>
>>
>
>
> --
> You received this message because you are subscribed to the Google
> Groups "mnemosyne-proj-users" group.
> To unsubscribe from this group and stop receiving emails from it,
> send an email to mnemosyne-proj-u...@googlegroups.com.
> To post to this group, send email to mnemosyne-...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/mnemosyne-proj-users/7a55af4e-371c-4cce-be22-48f6109d50ba%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages