Here is a 1.2Mb ASCII text file containing the text of Herman Melville's Moby-Dick; or, The Whale. Your task is to write a program or function (or class, etc. -- see below) which will be given this file one character at a time, and at each step must guess the next character.

This is code-challenge. Your score will be

2*L + E

where L is the size of your submission in bytes, and E is the number of characters it guesses incorrectly. The lowest score wins.

Further particulars

Your submission will be a program or function (etc.) that will be called or invoked multiple times. (1215235 times to be exact.) When it is called for the n^th time it will be given the n^th character of whale.txt and it must output its guess for the (n+1)^th character. The E component of its score will be the total number of characters that it guesses incorrectly.

Most submissions will need to store some state in between invocations, so that they can track how many times they have been called and what the previous inputs were. You can do this by writing to an external file, by using static or global variables, by submitting a class rather than a function, using a state monad, or whatever else works for your language. Your submission must include any code required to initialise its state before the first invocation.

Your program should run deterministically, so that it always makes the same guesses given the same input (and hence always gets the same score).

Your answer must include not only your submission, but also the code you used to calculate the E part of its score. This need not be written in the same language as your submission, and will not be counted towards its byte count. You are encouraged to make it readable.

Regarding the interface between your submission and this score-calculating program, anything is fine, as long as your program always gives one byte of output before receiving its next byte of input. (So, for example, you can't just pass it a string containing all of the input and get a string back containing all of the output.)

You must actually run your test program and calculate/verify your score before submitting your entry. If your submission runs too slowly for you to verify its score then it is not qualified to compete, even if you know what its score would be in principle.

The L component of your score will be calculated according to the usual rules for code golf challenges. If your submission will contain multiple files, please take note of the rules on scoring and directory structure in that case. Any data that your code uses must be included in your L score.

You may import existing libraries but may not load any other external files, and your code may not access the whale.txt file in any way other than described above. You may not load any pre-trained neural networks or other sources of statistical data. (It's fine to use neural networks, but you have to include the weight data in your submission and count it towards your byte count.) If for some reason your language or libraries include a feature that provides some or all of the text of Moby Dick, you may not use that feature. Aside from that you can use any other built-in or library features that you like, including ones relating to text processing, prediction or compression, as long as they're part of your language or its standard libraries. For more exotic, specialised routines that include sources of statistical data, you would have to implement them yourself and include them in your byte count.

It is likely that some submissions will include components that are themselves generated by code. If this is the case, please include in your answer the code that was used to produce them, and explain how it works. (As long as this code is not needed to run your submission it will not be included in your byte count.)